Join me and 2000 other AI engineers and builders at AI Dev in New York City, Nov 30-Dec 1!
A LinkedIn study found that over July-September, the job title AI Engineer has been one of the 10 fastest-growing occupations in seven major economies. AI engineering has become a core skill for all software work, and AI Dev is designed to help you develop it.
Our sessions give you hands-on practice with:
- Coding agents and harness design: Harness tuning, repository context, autonomous coding
- Agent memory and knowledge graphs: GraphRAG, vector search, persistent memory, hybrid retrieval
- Agents in production: Evals, observability, simulated testing, browser and voice agents
@ylecun and I will also talk about what’s next in world models and physical AI.
I hope to see you there!
Tickets: ai-dev.deeplearning.ai
Andrew Ng
Andrew Yan-Tak Ng (Chinese: 吳恩達; born April 18, 1976) is a British-American computer scientist and technology entrepreneur focusing on machine learning and artificial intelligence (AI). Ng was a cofounder and head of Google Brain and was the former Chief Scientist at Baidu. More on Wikipedia
-
Andrew Ng
@AndrewYNg
-
Andrew Ng
@AndrewYNg
Workera is being acquired by Pearson! It's been a privilege to support CEO @kiankatan and the whole Workera team through this journey. Back in 2019, Kian had the insight that rigorous skills measurements would be important. With recent advances in AI, this is now more true than ever. Workera's technologies for rigorously measuring people's skills in different tasks and job roles are helping many businesses understand where their employees are strong and where there are areas for development.
With Workera and Pearson combining forces, I'm delighted that under @omarabbosh and Kian's leadership. Workera now has the potential to support an even larger number of people and businesses. Thank you, Kian and the Workera team for having me serve as the company's chairman, for launching Workera out of deeplearning.ai and AI Fund, for your friendship over the years, and for your important work advancing assessments.
kian
@kiankatan
I have some big news to share. Workera is being acquired by Pearson!
Over six years ago, I was teaching at Stanford and thinking about a simple question: what if we could understand everyone's skills as precisely as the best teachers understand their students?
I believed it could lead to a more meritocratic world. People could be recognized for what they can actually do, not just their credentials or network. They could understand their strengths and gaps, and rapidly develop the skills they need next. Organizations could discover talent they might otherwise overlook and manage their workforce with trusted skills data.
What felt like a dream at the time is now a reality. Workera brought together experts in AI, psychometrics, and enterprise execution to build AI systems that reinvent how skills are measured. Our team pioneered AI-native skills intelligence, agent-led multimodal assessments, and even ambient skill measurement. We've established skills benchmarks across organizations, industries, and roles.
And this mission feels more important today than ever! AI is changing work as we speak. Some roles are disappearing, new ones are emerging, and we need to help billions of people develop new skills and navigate what comes next.
When I first spoke with @omarabbosh, it became clear that our companies shared the same mission. Pearson has helped generations of people learn and prove what they know. If you're reading this, there's a good chance you've taken a Pearson assessment, learned from their educational materials, earned a professional credential through them, read their psychometrics research, or benefited from their enterprise products in many other ways.
Bringing together Workera's technology and AI talent with Pearson’s global scale and deep expertise in learning and assessment means we can pursue our mission at a scale we could only imagine on our own.
To our customers and partners, thank you for believing in us. Expect even more innovations coming out of Workera and Pearson.
To the Workera team, I’m incredibly proud of what you've built, and your continued dedication to our beautiful mission. To our board and our chairman @AndrewYNg, thank you for your belief, support, and mentorship.
To everyone, we have big plans for this next chapter, so please stay tuned.
We're just getting started! 😊 -
Andrew Ng
@AndrewYNg
The OpenAI-Hugging Face hack was enabled by weak sandboxing. It is great that Nvidia is releasing open source tools for sandboxing AI agents. OpenWorker, our open-source agent harness supporting cybersecurity workflows, is proud to support this.
A sandbox gives an agent limited permissions. OpenWorker is building on Nvidia OpenShell and will support running each agent's commands inside a sandbox. Only the files relevant to the task go in. Secret API keys, your web browser login credentials, the ability to access arbitrary websites, are inaccessible to the agent by default. These restrictions are implemented in deterministic code rather than by prompting an LLM, which can make mistakes or be susceptible to prompt injections. Further, all actions are logged for monitoring and audit.
I'm grateful for @JensenHuang's leadership making AI agents more secure. OpenWorker (which @rohitcprasad and I are working on) will continue to improve security for agents.
Jensen Huang
@JensenHuang
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry.
Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to come.
But its full promise can only be realized when people have confidence that AI is being built to be safe and deployed with wisdom and responsibility.
This is bigger than a single product. It's the beginning of an open ecosystem to build the trust layer for safe agent systems.
Together, we are building the foundation of the AI economy.
Trust and innovation are not in conflict. Safety is how trust is earned. We must build not only the most capable AI, but the most trusted AI, so that this extraordinary technology can realize its enormous promise for the world. nvda.ws/4hcoq7m -
Andrew Ng reposted
Elon Musk
@elonmusk
Interesting perspective
Andrew Ng
@AndrewYNg
The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.
I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.
First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.
The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.
Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.
I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.
Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)
Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.
Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.
[Original text (with links): www.deeplearning.ai/the-batch/issue-371 ] -
Andrew Ng
@AndrewYNg
The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.
I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.
First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.
The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.
Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.
I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.
Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)
Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.
Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building.
[Original text (with links): www.deeplearning.ai/the-batch/issue-371 ] -
Andrew Ng
@AndrewYNg
With AI Engineering skills, you actively shape the build: You influence what gets built, and drive the build loop. Here're key skills to do this. x.com/i/article/2098450134883594240
-
Andrew Ng
@AndrewYNg
The most important skills for using AI coding agents effectively. Presenting the AI Engineering Skills Map for using coding agents. x.com/i/article/2095882148670832640
-
Andrew Ng
@AndrewYNg
How have software engineering fundamentals changed with agentic coding? Here is our AI Engineering Skills map for software engineering fundamentals. x.com/i/article/2093384274372419585
-
Andrew Ng
@AndrewYNg
OpenWorker -- an open source agent that doesn't just chat but completes tasks on your laptop -- just released a new version with many features for security workflows.
After our initial release, many users found it especially useful for cybersecurity. Attackers are already using AI; OpenWorker is committed to giving defenders the same leverage. Running an agent requires both (i) A model and (ii) A harness (the software around the model). Because the OpenWorker harness is fully open source, security teams can audit it to make sure we haven't built any backdoors that exfiltrate your code and data to some company or even a foreign adversary.
OpenWorker now comes with built-in cybersecurity agents for (i) Scanning your code for vulnerabilities. (ii) Scanning dependencies for supply chain injections. (iii) Checking your cloud security configuration for attack surfaces. This enables developers to do much more security work before deployment (part of what's called the "shift left" movement).
You choose the model: you can run open weight models fully locally so sensitive code never leaves your machine. This helps with legitimate security work (like reproducing a known exploit to defend against it) that can trigger refusals in leading closed models. Or use your ChatGPT subscription, or stealth preview models like Ox Alpha, or any model via API key.
Thanks also to all the open source contributors!
Join work with @rohitcprasad so please follow him too to get more frequent updates.
Try it out: openworker.com
Code: github.com/andrewyng/openworker -
Andrew Ng
@AndrewYNg
In the fight to defend openness in AI, the Marin project is a precious demonstration of openness in model training, with open code, data, recipes, even experimental results. Releasing AI research openly used to be the norm; I'm grateful for @percyliang's open lab approach.
Percy Liang
@percyliang
🚢 Marin 535B-A23B started training this week! As usual, the whole process is open.
Voyage plan: pretraining (80%) + midtraining (20%) on 18.75T tokens on 11 x GB200 NVL72 for ~3 months (2.7e24 FLOPs). Post-training will follow.
Before kicking off the run, we trained a 4-rung scaling ladder from 1.6B-A61M (48B tokens) to 27.7B-A1.2B (926B tokens) to debug issues, and to make a forecast of our hero run. This is by far our biggest run, so definitely expecting the unexpected. -
Andrew Ng
@AndrewYNg
The most important skills in Building and Deploying AI Applications. x.com/i/article/2090836273036763142
-
Andrew Ng
@AndrewYNg
New: A map of the most important skills in AI Engineering. x.com/i/article/2088296780983107584
-
Andrew Ng
@AndrewYNg
Thank you Mark, Alex and the whole Meta team for your contributions to open weight AI.
Mark Zuckerberg
@finkd
Today we're also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally. Soon we'll also release the weights for Muse Spark 1.2, our latest foundation model. Meta is a strong supporter of open source and I'm proud of these releases. Congrats to @alexandr_wang and the MSL team for all your great work on these models.
-
Andrew Ng
@AndrewYNg
Congratulations @JeffDean, @Sanjay_Ghemawat,
@OriolVinyalsML and @quocleix! Knowing you guys, this will be amazing. Cheering you on and wishing you well on your important work!
Jeff Dean
@JeffDean
Announcing Discovery Loop!
I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor.
♾
Learn more at: www.discoveryloop.com -
Andrew Ng
@AndrewYNg
Fifteen years ago, @Coursera and online courses changed education. It worked better than almost anyone expected, expanding access by opening up where you can learn. But how you learn remains largely the same as it has for centuries: it is still one-size-fits-all, taught the same way to each person who shows up.
We now have an opportunity to change how learning happens. With advances in AI, we can now build a custom learning guide for each person. We will turn learning from one‑to‑many to one‑to‑one. I'm starting LearnVector to invent this next generation of learning. We are starting with a $100M investment from Coursera, and plan to collaborate closely with Coursera and Udemy.
Good learning needs much more than just a chatbot. Research shows that chatbots without guardrails harm learning. They help complete tasks and enable students to do better on homework. But cognitive offloading to a chatbot results in them being less skilled. And, you cannot always trust what a chatbot tells you.
In contrast, LearnVector will plan a path with you, adapt to how you learn, and patiently stay with you until you’ve mastered new skills.
One thing has not changed in all this time. People want learning they can trust: material that is accurate, relevant, and worth the effort you put into it. Anything less wastes the most valuable thing a learner has: time. Coursera has a trusted library of materials from authoritative sources. LearnVector plans to work with Coursera to bring this trustworthy learning to everyone. I'm grateful to Greg Hart and the entire Coursera team for supporting LearnVector.
I look forward to working with our talented team to change how we learn, and accelerate human development.
learnvector.ai -
Andrew Ng
@AndrewYNg
Good move by @JensenHuang. The Nvidia letter is well written and worth reading. As we saw with the OpenAI-Hugging Face hack, we need open models and harnesses for defense.
Lets stop believing the PR that closed models are safer. - that's just regulatory capture.
Jensen Huang
@JensenHuang
Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.
During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.
That’s why we created the Open Secure AI Alliance.
NVIDIA
@nvidia
AI security advances when the industry builds in the open, together.
We're introducing the Open Secure AI Alliance with industry leaders to develop new techniques and tools to safeguard software and agents.
By sharing models, tooling and research in the open, we can broaden the community of defenders.
Learn more about the founding members’ contributions: nvda.ws/4pD8Fc5 -
Julian Schrittwieser
@Mononofu
I’m so excited that @JensenHuang is a believer in open source now, looking forward to the CUDA and GPU driver open source release!
Andrew Ng
@AndrewYNg
This is a false equivalence. Everyone has the right to keep their code private. The problem is when someone tries to stop OTHERS from open sourcing.
-
Andrew Ng
@AndrewYNg
Announcing OpenWorker! An open-source agent that doesn't just chat with you, but delivers finished work -- like hand you a polished document, send a slack message, or update a calendar entry.
Ask it to prepare a customer brief, untangle your calendar, draft a report, or triage a Slack alert. It works across your files and everyday tools, produces the deliverable, and checks in before doing anything consequential.
OpenWorker runs on your Mac, with Windows support coming soon. It does not lock you into any one model. Bring your own API key and run it with GPT 5.6 Sol, Claude Fable, Gemini 3.6, an open weight model (like Kimi, GLM, DeepSeek, Inkling), or Ollama to keep your data local. Your data does not leave your machine except through an LLM provider and integrations that you choose.
@rohitcprasad and I are building OpenWorker because AI coworkers are an important way to get work done, and we want there to be an open, privacy-preserving, model-independent option. Check it out and let us know what you think!
Try it out: openworker.com (requires your own API key)
Source code: github.com/andrewyng/openworker -
Andrew Ng
@AndrewYNg
New course: Build LLM applications that respond to user requests quickly by running on hardware designed for fast inference. This short course was built with @Cerebras and taught by @zhennydez, @duerr_seb, and @MilksandMatcha.
When a model generates text, much of the time is spent moving its weights out of memory and into the compute units. Inference-optimized hardware minimizes that movement, making token generation several times faster than on a typical GPU setup. In this course, the hardware you'll use is Cerebras' Wafer-Scale Engine, which is designed for fast inference by keeping the model's weights close to the compute units.
Fast inference makes lengthy agentic workflows go faster, and also unlocks latency-sensitive, real-time applications like live translation and voice agents.
Skills you'll gain:
- Compare how GPUs, TPUs, and Cerebras' Wafer-Scale Engine each handle the memory-to-compute bottleneck
- Build real-time applications powered by fast inference, including personalizing a webpage and running a multi-step workflow to analyze market signals
- Adopt concrete habits for agentic coding with fast inference, keeping your sessions focused and steering the model more effectively
My teams use Cerebras for several applications that are latency sensitive. Join and build LLM applications that respond quickly:
www.deeplearning.ai/courses/fast-llm-inference-with-cerebrasThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Adam Thierer had written the landmark book on "Permissionless Innovation." We get the best ideas when we don't have to ask the government in advance for permission to invent. Protecting open source AI is now a critical part of ensuring this.
Adam Thierer
@AdamThierer
We stand at a critical crossroads in the debate over AI governance in the United States, and it feels like we are inching closer to a very serious battle over whether or not open source models will even be allowed in an environment where a new de facto licensing regime has been taking shape.
Lacking formal congressional statutory frameworks or clear administration rules (like the diffusion rule revision), we appear to be left with a sporadic, arbitrary, non-transparent process for model review. The fiction of “voluntary” agreements hangs over this debate, and some large model developers are already showing an incredible willingness to bend over backwards to accommodate national security-related officials / orders that the rest of us are not privy to. It's a very opaque process. And those model developers are expected to play ball with those officials, or else their models get pulled from the market or held up for long periods. Or they will lose any government procurement contracts they have. There is nothing “voluntary” about it when that Sword of Damocles hangs in the room.
As this mess worsens, at some point the question of how to handle open source models will come into sharper focus because it will have to. I've even heard some rumors lately that something may be coming from the admin on this front to address this.
Needless to say, if this informal new AI model review regime expands and takes on more pre-vetting characteristics / requirements, it is hard to see how open source players could comply with such quasi-licensing of AI models. Specifically, if this ambiguous new regime is accompanied by a general presumption of ‘restrict-until-permitted,’ then that would spell doom for open source. That is a very dark path for our country.
Worse yet, of course, would be a move by national security officials to more directly restrict open source models and capabilities. If that happens, then we would be right back in the thick of a Clipper Chip-like battle along the lines of what we saw in the late 1990s. That is a much darker path for America.
Meanwhile, open source developers have no “golden shares” or other goodies to offer the government to make their problems go away.
Let’s be clear: If our government takes the dark path, it will become the single most important battle over computational freedom of modern times. It is time for people to make a stand in defense of open source before it is too late. -
Andrew Ng
@AndrewYNg
“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code’s creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software. In this letter, I’d like to share my 3 key loops, shown in the image below, for building 0-to-1 products. These loops guide not just how I build software, but also how I decide what software to build.
Agentic coding loop: Given a product specification and optionally a set of evals (that is, a dataset against which to measure performance), we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification. This idea of closing the loop took off around the end of last year, and it has been a game changer in enabling coding agents to work longer productively without human intervention. For example, over the weekend, I was building an app for my daughter to practice typing, and my coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention.
The engineering loop executes quickly. Every few minutes, the coding agent might build and test a new version of the software. I hear frequently from developers who are finding new ways to engineer more effective engineering loops. This is an active area of invention!
Developer feedback loop: In this loop, a developer examines the current product and steers the coding agent to improve it. Last year, a lot of developers (including me) were acting as the QA (quality assurance) function for our coding agents, manually finding bugs and then asking the agent to fix them. But with coding agents much more able to test their own code, the amount of time we need to spend on this function has decreased significantly. This allows us to make higher-level product decisions, such as what key features to offer, where the UI needs improvement, and so on.
The developer-feedback loop operates over time intervals between tens of minutes and hours — that's how frequently a developer might review a product and give feedback. In the case of the typing app, I changed my mind a few times about the visual design, what cat costumes she can unlock as she learns (she loves cats), and the user flow for a grown-up to log in and steer the child's learning experience.
When a developer has a clear vision for what to build, it is still a lot of work to translate that vision into a specification for a coding agent to implement. Further, after the developer has seen an implementation, they might update (or perhaps clarify) the spec to steer it toward what they want. If you find that the system repeatedly runs into certain problems, building a set of evals for the agent becomes useful.
AI-native teams are increasingly using AI to help shape product direction, for example, automating the gathering and analysis of usage data, summarizing written and verbal customer feedback, or carrying out competitive analysis. However, for pretty much all the products I’m involved in, I see humans as having a significant context advantage over current AI systems — we know a lot more than the AI system about the users and the context the product has to operate in — and thus humans play a critical role. Many people describe this human contribution as “taste,” but I prefer to think of it as humans having a context advantage, since that gives us a clearer path to helping AI systems get better. This also speaks to why this step can’t be automated: So long as the human knows something the AI does not, human-in-the-loop is needed to to inject that knowledge into the system.
External feedback loop: This includes a wide range of tactics like asking a few friends for feedback, launching to alpha testers, or putting the code into production with A/B testing. These tactics are usually slow, rarely taking less than hours and sometimes taking days or even weeks. This data informs the developer vision, which in turn continues to drive the detailed product spec, which in turn drives the coding agent.
With coding agents speeding up software development, more engineers are starting to play a partial product management role. For many engineers who are growing into this role, the hardest part is shaping the product vision and striking a balance between building (bridging the gap between vision and spec) and getting user feedback to evolve the vision. It is important to do both!
I will write more about how to do this in future posts, but for now, I find it encouraging that engineers are playing an expanded role (just as product managers and designers now do more engineering).
[Original text: The Batch] -
Andrew Ng
@AndrewYNg
Over the last two weeks, both the U.S. Government and Anthropic took significant actions that demonstrated their power to control access to AI by restricting what others can do with frontier models. This has been one of those moments that, once seen, will be hard to unsee, and it is significantly accelerating many businesses’ and nation states’ efforts to ensure reliable access to AI that no one else can terminate.
Anthropic first released Claude Fable 5, a version of its Mythos model with additional guardrails, including some restrictions that seem well justified on safety grounds (such as limitations on applying it to hacking, bioweapons, and so forth). However, it also restricted developers’ ability to use it to build competing LLM technology. This move was concerning, given that the whole AI community, including Anthropic, has benefitted tremendously from open research — indeed, the AI revolution was kicked off by my former team (Google Brain) freely publishing the Transformers paper!
Imagine if Microsoft’s terms of use barred anyone from using their tools to build competitive software, or if Google barred using it to search for information to work on competing search engines. Anthropic’s argument that it was unsafe for others to be able to make advances in AI also rang hollow. Initially, Anthropic silently degraded Fable 5’s performance for users detected to be working on LLM research through invisible interventions that weakened the model’s outputs without notifying the user. After significant backlash, it walked back this decision and decided to be transparent when it did this, but it still refuses to use its latest capabilities to help AI researchers.
This move represents a raw demonstration of power by Anthropic. It has used “safety” arguments to hinder potential competitors. Platforms succeed when they are viewed as stable, reliable partners that one can build on. The sudden rule changes by Anthropic (including a mandatory 30 day data retention policy for Fable usage) have made developers wonder about the stability of building on any one proprietary LLM provider, not just Anthropic.
The U.S. Government then shortly followed with an even greater demonstration of power. It used the Commerce Department’s authority to regulate technologies that may be national security threats to restrict exports of Mythos and Fable, requiring a license for use by any foreign national, whether inside or outside of the U.S., including employees of Anthropic. This led Anthropic to disable access to Fable to all users worldwide.
Sam Altman pointed out, referring to Anthropic, “It is clearly incredible marketing to say, ‘We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million.’” But when one engages in this type of fear-based marketing, it increases the odds that the U.S. Government will agree with you and slap export controls on the bomb you say you have built.
To be clear, I don't think Anthropic has built anything like a bomb, and I don't think export controls on Fable are appropriate.
However, following the U.S. Government making this move, many nations, including U.S. allies, saw how the U.S. can suddenly yank their access to AI models. In many capitals around the world, this has spurred discussions on AI sovereignty and how others can ensure uninterrupted access to this critical technology.
For decades, many nations were comfortable having many parts of their supply chain rely on the U.S., China, and other major producers. Once a nation issues a threat, or takes action, to limit other nations’ access, other nations will rationally try to secure alternatives. For decades, semiconductor manufacturing in China made slow progress; once the U.S. moved to limit China’s access, China’s efforts kicked into high gear. Similarly, once China threatened U.S. access to rare earth minerals, U.S. efforts to secure alternatives accelerated. Now that it has become crystal clear that private U.S. companies and the U.S. government can limit, in short order, other nations’ access to frontier AI models, the incentive of others to invest more in alternatives like open source grows significantly. Of course, training frontier models is not easy, so it remains to be seen how successful they are, but we have crossed the rubicon.
Satya Nadella wrote an essay about the importance of building a healthy ecosystem on top of frontier AI technology. I heartily agree with him, and hope this week’s events will ultimately prove to be constructive steps toward this.
I hope we can build a more free, more open world, where research is freely shared, and laws and societal norms shape a level playing field that allows everyone to make progress. A silver lining of the events of these past two weeks is now that everyone better realizes key points of instability of the current system, we can all work to create a more stable foundation.
[Original text: The Batch newsletter] -
Andrew Ng
@AndrewYNg
New course: Add voice to your AI agents and applications, built with @VocalBridge (disclosure: an AI Fund portfolio company) and taught by its CEO @_ashwyn.
Voice applications historically required making a hard tradeoff: using fast voice-to-voice models that sacrifice reliability, or accurate speech-to-text pipelines that add latency. This course teaches you how to build voice agents that are both reliable and fast.
You'll build three types of voice-enabled applications: a voice-interactive game where voice commands and mouse clicks work together over a single channel, an agent that gains a voice in about 10 lines of code without touching its prompts or tools, and an agent that places outbound phone calls using a make_phone_call function.
Skills you'll gain:
- Add a voice layer to an existing agent without rewriting your prompts, RAG pipeline, or tools
- Give an agent the ability to place outbound calls and stream transcripts back live
- Set up voice evaluation to score calls, catch regressions, and improve quality before deployment
Join and add voice to your agents without overhauling your architecture:
www.deeplearning.ai/courses/voice-for-ai-agents-and-appli…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New course on serving LLMs efficiently -- how do you serve models to many concurrent users at low latency and reasonable cost? This short course is built with @RedHat and taught by @cedricclyburn.
Efficient LLM serving requires efficient memory management. A 70B-parameter model takes ~140 GB just to load the weights. On top of that, every active request needs its own chunk of GPU memory, the KV cache, to store the token context it has built up so far. In this course, you'll learn to reduce a model's memory footprint with quantization and serve it using vLLM, which handles many concurrent requests efficiently through smart memory management.
Skills you'll gain:
- Quantize a model and measure the accuracy tradeoff
- Serve a model with vLLM and watch it handle concurrent requests efficiently
- Benchmark your deployment and make informed tradeoffs between speed, cost, and accuracy
Join and learn to serve LLMs efficiently:
www.deeplearning.ai/courses/fast-and-efficient-llm-infere…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
One of the new, buzzy jobs in Silicon Valley is the AI Forward Deployed Engineer (FDE), an engineer who is embedded within a client organization to help customize solutions, such as building and tuning agentic workflows that suit the client’s particular needs. I’ve heard from people who are wondering anew about the FDE career path since OpenAI and Anthropic started building new teams to place FDEs within client organizations.
The rise of FDEs for AI workloads is one way AI is creating new jobs (and why the jobpolcalypse narrative of upcoming job market collapse is false -- there will be many AI and non-AI jobs). However, I believe there will be far more AI Engineer jobs than FDEs, as I explain below.
The FDE role was pioneered about two decades ago by Palantir, which sent engineers to government locations to work on secure, air-gapped networks. In addition to having good technical skills, FDEs need communication skills and sometimes business skills. For example, they may need to speak with clients to understand their needs, formulate a strategy to prioritize projects, explain complex technology, and respectfully push back if a client asks for something unrealistic. They’re enjoying a resurgence because of the amount of work involved in taking an off-the-shelf LLM and building it into a custom agentic workflow that fits particular business needs.
However, I believe the number of AI Engineer jobs will be far larger. A company might accept a few FDEs to be embedded within its organization. But most companies will want far more of their own employees working on their projects. While my organizations do hire FDEs, we hire far more AI Engineers! Also, a common client concern is that it is hard to find vendor-neutral FDEs — they are, after all, there to deeply integrate a particular vendor’s product into a company. In this moment when it’s hard to predict which AI service will be the best one in a year’s time, optionality (the ability to pick whatever vendor turns out to fit best in the future) is very valuable. In contrast, letting FDEs tightly bind a company’s processes significantly reduces optionality.
Right now, I see surging demand for AI Engineers who can build software applications using AI software components (like LLM prompting, agentic frameworks, evals, etc.) and effectively use AI coding agents (like Claude Code, Codex, Antigravity CLI, and OpenCode). As the AI Engineer role matures, I expect it to fragment into more specialized roles, like the generic Software Engineer role from decades ago fragmented into frontend, backend, mobile, data engineering, devops, and so on.
What will be the future, specialized AI engineering roles? I don’t know. Perhaps there will be AI FDEs, LLMOps Engineers, Evals Engineers, AI Data Engineers, Harness Engineers, and other roles we don’t have names for yet. But for now, I see a lot of AI engineers who are generalists create a lot of value. Skilled AI Engineers are in very high demand! As our field continues to mature over the coming decade, I look forward to new specializations within AI Engineering that create even more job opportunities.
[Original text: The Batch newsletter] -
Andrew Ng
@AndrewYNg
Memorial Day. Today we honor those who gave everything to defend our democracy. Democracy isn't guaranteed; it's a precious inheritance that requires our constant care. 🇺🇸
-
Andrew Ng
@AndrewYNg
The new White House policy requiring green card applicants to apply from outside the US is a capricious attack on legal immigration. It will hurt families, leave us with fewer doctors, teachers and scientists, and hurt American competitiveness in AI.
-
Andrew Ng
@AndrewYNg
Harvard University just voted to limit the number of A grades given in undergraduate classes to about 20% of the class. I’m not in favor of this. It deeply runs counter to how I believe education should be. We should hold a high bar, but also work mightily to support the success of 100% of learners, rather than a fraction.
Harvard’s administration took this step — over the objections of a large fraction of the student body — to counter grade inflation. Grade inflation is real: Many universities have been awarding A and B grades to ever larger fractions of students, and this has caused grade point averages (GPAs) to become less useful as signals of student skill. At the same time, we want students to succeed. The heart of the question is the role of educational institutions. Should our goal be:
- To help students succeed?
- To judge students?
Both of these have value. But my focus when working in education is almost entirely helping students succeed.
To me, it is clear that many people want to learn, to be empowered, to build skills that let them do new things! This is what we focus on at DeepLearningAI. This philosophy is also why my online courses (going back to my early online Stanford courses on Coursera) permitted an unlimited number of retries for graded assignments.
I believe in letting — and even encouraging — someone to redo something until they succeed. This is as opposed to standing in judgement of the fact they didn’t get it right the first time. Further, I want homework assignments to be designed primarily to help people practice and learn, rather than to judge their skill level. This is why I prefer to create “Practice Problems” and “Practice Labs” — questions that, when you think through them, help you to gain practice and reinforce what you know. As opposed to “Assessment Problems” designed primarily to judge skill.
But won’t Harvard’s move make GPAs more meaningful and help prospective employers identify strong candidates? Having hired a large number of people from Harvard and other institutions, I can say confidently that GPA is not an important signal. We have screening and interviewing processes that give far more accurate ways to figure out if someone is truly skilled. I do not need a wider spread in applicant GPA scores to figure out who's really good!
To be clear, there is also value in assessment. Even though standardized testing is much hated, high-quality tests like the SAT, ACT, GRE, TOEFL, etc. provide objective measures of ability in a domain. I find that most people want to learn and succeed. There are also people who want rigorous assessment (for example, to apply for school admissions), but this is a lesser need, and is not my focus when building educational products.
Harvard is often described as an “elite” educational institution. There are two ways to be elite: One option involves limiting enrollments, and then even among admitted students, cap the number of people that do well at 20%. I would rather pursue a different path: Set a high bar and teach elite, cutting-edge skills, but strive relentlessly to help everyone succeed. This way, eliteness is defined not by excluding people but by helping as many people as possible to be excellent.
[Original text: The Batch newsletter] -
Andrew Ng
@AndrewYNg
New course: Build AI agents that generate images and videos -- an under-explored frontier. A key to performance is having the agent evaluate its own output, and iterate to improve quality. This short course is built together with @googlecloudtech and taught by Katie Nguyen and Wafae Bakkali.
You'll learn three evaluation techniques and combine them in an agent: image-text similarity scoring to check the output matches the prompt, an LLM judge that scores against custom criteria like brand consistency, and structured rubrics that break a prompt into verifiable yes/no questions like "is the subject in the frame?" and "does the camera motion match?"
Skills you'll gain:
- Learn image and video prompt engineering
- Build an image agent that turns brand guidelines into UI mockups
- Build a video agent that plans multi-scene explainers and animates reference frames with synchronized audio
Join and build agents that create images and video!
www.deeplearning.ai/courses/ai-agents-for-image-and-video…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New course: Transformers in Practice. You'll get a practical view of how transformer-based LLMs work, so you can reason about their behavior, diagnose problems like slow inference, and make smarter decisions about deployment. This course is built in partnership with @AMD and taught by @realSharonZhou.
You'll see how transformers generate text one token at a time, how the model decides which earlier words matter most when predicting the next one, and how techniques like quantization speed up inference on GPUs. This is not a video-only course; interactive visualizations throughout let you play with these concepts and build intuition that sticks.
Skills you'll gain:
- Understand why LLMs hallucinate, and RAG and chain-of-thought shape what they generate
- Look inside the model to see how attention and layers combine to predict the next token
- Diagnose inference bottlenecks and learn the techniques that speed up transformers on GPUs
Join and understand what's really happening inside your LLMs: www.deeplearning.ai/courses/transformers-in-practiceThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
There will be no AI jobpocalypse.
The story that AI will lead to massive unemployment is stoking unnecessary fear. AI — like any other technology — does affect jobs, but telling overblown stories of large-scale unemployment is irresponsible and damaging. Let’s put a stop to it.
I’ve expressed skepticism about the jobpocalypse in previous posts. I’m glad to see that the popular press is now pushing back on this narrative. The image below features some recent headlines.
Software engineering is the sector most affected by AI tools, as coding agents race ahead. Yet hiring of software engineers remains strong! So while there are examples of AI taking away jobs, the trends strongly suggest the net job creation is vastly greater than the job destruction — just like earlier waves of technology. Further, despite all the exciting progress in AI, the U.S. unemployment rate remains a healthy 4.3%.
Why is the AI jobpocalypse narrative so popular? For one thing, frontier AI labs have a strong incentive to tell stories that make AI technology sound more powerful. At their most extreme, they promote science-fiction scenarios of AI “taking over” and causing human extinction. If a technology can replace many employees, surely that technology must be very valuable!
Also, a lot of SaaS software companies charge around $100-$1000 per user/year. But if an AI company can replace an employee who makes $100,000 — or make them 50% more productive — then charging even $10,000 starts to look reasonable. By anchoring not to typical SaaS prices but to salaries of employees, AI companies can charge a lot more.
Additionally, businesses have a strong incentive to talk about layoffs as if they were caused by AI. After all, talking about how they’re using AI to be far more productive with fewer staff makes them look smart. This is a better message than admitting they overhired during the pandemic when capital was abundant due to low interest rates and a massive government financial stimulus.
To be clear, I recognize that AI is causing a lot of people’s work to change. This is hard. This is stressful. (And to some, it can be fun.) I empathize with everyone affected. At the same time, this is very different from predicting a collapse of the job market.
Societies are capable of telling themselves stories for years that have little basis in reality and lead to poor society-wide decision making. For example, fears over nuclear plant safety led to under-investment in nuclear power. Fears of the “population bomb” in the 1960s led countries to implement harsh policies to reduce their populations. And worries about dietary fat led governments to promote unhealthy high-sugar diets for decades.
Now that mainstream media is openly skeptical about the jobpocalypse, I hope these stories will start to lose their teeth (much like fears of AI-driven human extinction have).
Contrary to the predictions of an AI jobpocalypse, I predict the opposite: There will be an AI jobapalooza! AI will lead to a lot more good AI engineering jobs, and I’m also optimistic about the future of the overall job market. What AI engineers do will be different from traditional software engineering, and many of these jobs will be in businesses other than traditional large employers of developers. In non-AI roles, too, the skills needed will change because of AI. That makes this a good time to encourage more people to become proficient in AI, and make sure they’re ready for the different but plentiful jobs of the future!
[Original text in The Batch newsletter.] -
Andrew Ng
@AndrewYNg
I'm delighted that @coursera and @udemy have come together as one company to serve learners.
Both Coursera and Udemy were founded with the belief that access to high-quality education changes lives. Over the years, both companies have advanced this goal, creating opportunities for individuals, organizations, and communities around the world.
That role is even more important now, as AI is changing the nature of work and increasing the need for continuous learning. Helping people build job-relevant skills will be critical to how we create a better world.
By combining the strengths of both companies, we can better serve this need. We bring together a broader range of learning content, trusted instructors and educators, and engaging learning experiences. This creates new opportunities to make learning more personalized, more applied, and more accessible at scale.
I’m excited to serve as Chairman of the combined company, working alongside Greg Hart and the leadership team. There is a strong foundation in both organizations, and I look forward to what the teams will build together to expand access opportunity globally.
Learn more: blog.coursera.org/coursera-and-udemy-are-now-one-company-… -
Andrew Ng
@AndrewYNg
New course: Build agents that respond to users with not only plaintext, but custom UIs like charts, forms, and whiteboards, generated on demand and displayed right in the chat. This short course is built in partnership with @CopilotKit and taught by @ataiiam, co-founder of CopilotKit.
You'll learn three approaches: Your agent can pick from custom components you build, like charts and forms. It can compose new layouts from a set of building blocks you provide, like rows, cards, and text. Or it can incorporate existing third-party apps, like a whiteboard or a calendar, right inside the conversation.
Skills you’ll gain:
- Build agents that render custom components like charts and forms on demand
- Build an app where the agent and user collaborate on shared data, beyond just the chat window
- Place third-party apps like maps, calendars, and whiteboards right in your interface
Join and build agents that give users something to see and act on! www.deeplearning.ai/short-courses/build-interactive-agent…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Coding agents are accelerating different types of software work to different degrees. When we architect teams, understanding these distinctions helps us to have realistic expectations. Listing functions from most accelerated to least, my order is: frontend development, backend, infrastructure, and research.
Frontend development — say, building a web page to serve descriptions of products for an ecommerce site — is dramatically sped up because coding agents are fluent in popular frontend languages like TypeScript and JavaScript and frameworks like React and Angular. Additionally, by examining what they have built by operating a web browser, coding agents are now very good at closing the loop and iterating on their own implementations. Granted, LLMs today are still weak at visual design, but given a design (or if a polished design isn’t important), the implementation is fast!
Backend development — say, building APIs to respond to queries requesting product data — is harder. It takes more work by human developers to steer modern models to think through corner cases that might lead to subtle bugs or security flaws. Further, a backend bug can lead to non-intuitive downstream effects like a corrupted database that occasionally returns incorrect results, which can be harder to debug than a typical frontend bug. Finally, although database migrations can be easier with coding agents, they’re still hard and need to be handled carefully to prevent data loss. While backend development is much faster with coding agents, they accelerate it less, and skilled developers still design and implement far better backends than inexperienced ones who use coding agents.
Infrastructure. Agents are even less effective in tasks like scaling an ecommerce site to 10K active uses while maintaining 99.99% reliability. LLMs' knowledge is still relatively limited with respect to infrastructure and the complex tradeoffs good engineers must make, so I rarely trust them for critical infra decisions. Building good infrastructure often requires a period of testing and experimentation, and coding agents can help with that, but ultimately that’s a significant bottleneck where fast AI coding does not help much. Lastly, finding infrastructure bugs — say, a subtle network misconfiguration — can be incredibly difficult and requires deep engineering expertise. Thus, I’ve found that coding agents accelerate critical infrastructure even less than backend development.
Research. Coding agents accelerate research work even less. Research involves thinking through new ideas, formulating hypotheses, running experiments, interpreting them to potentially modify the hypotheses, and iterating until we reach conclusions. Coding agents can speed up the pace at which we can write research code. (I also use coding agents to help me orchestrate and keep track of experiments, which makes it easier for a single researcher to manage more experiments.) But there is a lot of work in research other than coding, and today’s agents help with research only marginally.
Categorizing software work into frontend, backend, infra, and research is an extreme simplification, but having a simple mental model for how much different tasks have sped up has been useful for how I organize software teams. For example, I now ask front-end teams to implement products dramatically faster than a year ago, but my expectations for research teams have not shifted nearly as much.
I am fascinated by how to organize software teams to use coding agents to achieve speed, and will keep sharing my findings in future posts.
[Original text: www.deeplearning.ai/the-batch/issue-350/ ] -
Andrew Ng
@AndrewYNg
How we prompt AI is very different in 2026 than 2022 when ChatGPT came out.
I'm teaching a new course, AI Prompting for Everyone, to help you become an AI power user — whatever your current skill level.
It covers skills that apply across ChatGPT, Gemini, Claude, and other AI tools. How to use deep research mode for well-researched reports on complex questions. How to give AI the right context, including more documents and images than most people realize you can provide. When to ask AI to think hard for several minutes on important decisions like what car to buy, what to study, or what job to take. And how to use AI to generate images, analyze data, and build simple games and websites.
I also cover intuitions about how these models work under the hood, so you know when to trust an answer and when not to.
Along the way, you'll see flying squirrels, a creativity test, some of my old family photos, and fireworks.
Join me at deeplearning.ai/courses/ai-prompting-for-everyoneThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI-native software engineering teams operate very differently than traditional teams. The obvious difference is that AI-native teams use coding agents to build products much faster, but this leads to many other changes in how we operate. For example, some great engineers now play broader roles than just writing code. They are partly product managers, designers, sometimes marketers. Further, small teams who work in the same office, where they can communicate face-to-face, can move incredibly quickly.
Because we can now build fast, a greater fraction of time must be spent deciding what to build. To deal with this project-management bottleneck, some teams are pushing engineer:product manager (PM) some teams are pushing engineer:product manager (PM) ratios downward from, say, 8:1 to as low as 1:1. But we can do even better: If we have one PM who decides what to build and one engineer who builds it, the communication between them becomes a bottleneck. This is why the fastest-moving teams I see tend to have engineers who know how to do some product work (and, optionally, some PMs who know how to do some engineering work). When an engineer understands users and can make decisions on what to build and build it directly, they can execute incredibly quickly.
I’ve seen engineers successfully expand their roles to including making product decisions, and PMs expand their roles to building software. The tech industry has more engineers than PMs, but both are promising paths. If you are an engineer, you’ll find it useful to learn some product management skills, and if you’re a PM, please learn to build!
Looking beyond the product-management bottleneck, I also see bottlenecks in design, marketing, legal compliance, and much more. When we speed up coding 10x or 100x, everything else becomes slow in comparison. For example, some of my teams have built great features so quickly that the marketing organization was left scrambling to figure out how to communicate them to users — a marketing bottleneck. Or when a team can build software in a day that the legal department needs a week to review, that’s a legal compliance bottleneck. In this way, agentic coding isn’t just changing the workflow of software engineering, it’s also changing all the teams around it.
When smaller, AI-enabled teams can get more done, generalists excel. Traditional companies need to pull together people from many specialties — engineering, product management, design, marketing, legal, etc. — to execute projects and create value. This has resulted in large teams of specialists who work together. But if a team of 2 persons is to get work done that require 5 different specialities, then some of those individuals must play roles outside a single speciality. In some small teams, individuals do have deep specializations. For example, one might be a great engineer and another a great PM. But they also understand the other key functions needed to move a project forward, and can jump into thinking through other kinds of problems as needed. Of course, proficiency with AI tools is a big help, since it helps us to think through problems that involve different roles.
Even in a two-person team, to move fast, communication bottlenecks also must be minimized. This is why I value teams that work in the same location. Remote teams can perform well too, but the highest speed is achieved by having everyone in the room, able to communicate instantaneously to solve problems.
This post focuses on AI-native teams with around 2-10 persons, but not everything can be done by a small team. I'll address the coordination of larger teams in the future.
I realize these shifts to job roles are tough to navigate for many people. At the same time, I am encouraged that individuals and small teams who are willing to learn the relevant skills are now able to get far more done than was possible before. This is the golden age of learning and building!
[Original text: www.deeplearning.ai/the-batch/issue-349/ ] -
Andrew Ng
@AndrewYNg
New course: Spec-Driven Development with Coding Agents, built in partnership with @jetbrains, and taught by @paulweveritt.
Vibe coding is fast, but often produces code that doesn't match what you asked for. This short course teaches you spec-driven development: write a detailed spec defining what to build, and work with your coding agent to implement it. Many of the best developers already build this way.
A spec lets you control large code changes with a few words, preserve context across agent sessions, and stay in control as your project grows in complexity.
Skills you'll gain:
- Write a detailed specification to define your mission, tech stack, and roadmap, giving your agent the context it needs from the start
- Plan, implement, and validate features in iterative loops using a spec as your agent's guide
- Apply the same repeatable workflow to both new and legacy codebases
- Package your workflow into a portable agent skill that works across agents and IDEs
Join and write specs that keep your coding agent on track!
www.deeplearning.ai/short-courses/spec-driven-developmentThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I'm excited about voice as a UI layer for existing visual applications — where speech and screen update together. This goes well beyond voice-only use cases like call center automation.
The barrier has been a hard technical tradeoff: low-latency voice models lack reliability, while agentic pipelines (speech-to-text → LLM → text-to-speech) are intelligent but too slow for conversation. Ashwyn Sharma and team at Vocal Bridge (an AI Fund portfolio company) address this with a dual-agent architecture: a foreground agent for real-time conversation, a background agent for reasoning, guardrails, and tool calls.
I used Vocal Bridge to add voice to a math-quiz app I'd built for my daughter; this took less than an hour with Claude Code. She speaks her answers, the app responds verbally and updates the questions and animations on screen.
Only a tiny fraction of developers have ever built a voice app. If you'd like to try building one, check out Vocal Bridge for free: vocalbridgeai.com -
Andrew Ng
@AndrewYNg
As AI agents accelerate coding, what is the future of software engineering? Some trends are clear, such as the Product Management Bottleneck, referring to the idea that we are more constrained by deciding what to build rather than the actual building. But many implications, like AI’s impact on the job market, how software teams will be organized, and more, are still being sorted out.
The theme of our AI Developer Conference on April 28-29 in San Francisco is The Future of Software Engineering. I look forward to speaking about this topic there, hearing from other speakers on this theme, and chatting with attendees about it. We’re shaping the future, and I hope you will join me there!
It is currently trendy in some technology and policy circles to forecast massive job losses due to AI. Even if they have not yet materialized, these losses certainly must be just over the horizon! I have a contrarian view that the AI jobpocalypse — the notion that AI will lead to massive unemployment, perhaps even rioting in the streets — won’t be nearly as bad as dire forecasts by pundits, especially pundits who are trying to paint a picture of how powerful their AI technology is.
Among professions, AI is accelerating software engineering most, given the rise of coding agents. According to a new report by Citadel Research, software engineering job postings are rising rapidly. So if software engineering is a harbinger of the impact AI will have on other professions, this expansion of software engineering jobs is encouraging.
Yes, fresh college graduates are having a hard time finding jobs. And yes, there have been layoffs that CEOs have attributed to AI, even if a large fraction of this was “AI washing,” where businesses choose to attribute layoffs to AI, even though AI has not changed their internal operations much yet. And yes, there is a subset of job roles, such as call center operator, that are more heavily impacted. Many people are feeling significant job insecurity, and I feel for everyone struggling with employment, whether or not the cause is AI-related. And many other factors, such as over-hiring during the pandemic and high interest rates, have contributed to the slowdown in the labor market, and the notion that AI is leading to unemployment is oversimplified.
In software engineering, I see a lot of exciting work ahead to adapt our workflows. It is already clear that: (i) As AI makes coding easier, a lot more people will be doing it. (ii) Writing code by hand and even reading (generated) code is not that important, because we can ask an LLM about the code and operate at a higher level than the raw syntax (although how high we can or should go is rapidly changing). (iii) There will be a lot more custom applications, because now it’s economical to write software for smaller and smaller audiences. (iv) Deciding what to build, more than the actual building, is becoming a bottleneck. (v) The cost of paying down technical debt is decreasing (since AI can refactor for you).
At the same time, there are also a lot of open questions for our profession, such as:
- In the future, what will be the key skills of a senior software engineer? And for junior levels, what should be the new Computer Science curriculum?
- If everyone can build features, what skills, strategies, or resources create competitive advantage for individuals and for businesses?
- What are the new building blocks (libraries, SDKs, etc.) of software? How do we organize coding agents to create software?
- What should a software team look like? For example, how many engineers, product managers, designers, and so on. What tooling do we need to manage their workflow?
- How do AI agents change the workflow of machine learning engineers and data scientists? For example, how can we use agents to accelerate exploring data, identifying hypotheses, and testing them?
I’m excited to explore these and other questions about the future of software engineering at AI Dev. I expect this to be an exciting event. Please join us!
[Original text: The Batch newsletter.]
ai-dev.deeplearning.ai -
Andrew Ng
@AndrewYNg
New course: Efficient Inference with SGLang: Text and Image Generation, built in partnership with LMSys @lmsysorg and RadixArk @radixark, and taught by Richard Chen @richardczl, a Member of Technical Staff at RadixArk.
Running LLMs in production is expensive, and much of that cost comes from redundant computation. This short course teaches you to eliminate that waste using SGLang, an open-source inference framework that caches computation already done and reuses it across future requests.
When ten users share the same system prompt, SGLang processes it once, not ten times. The speedups compound quickly, especially when there's a lot of shared context across requests.
Skills you'll gain:
- Implement a KV cache from scratch to eliminate redundant computation within a single request
- Scale caching across users and requests with RadixAttention, so shared context is only processed once
- Accelerate image generation with diffusion models using SGLang's caching and multi-GPU parallelism
Join and learn to make LLM inference faster and more cost-efficient at scale!
www.deeplearning.ai/short-courses/efficient-inference-wit…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
The anti-AI coalition continues to maneuver to find arguments to slow down AI progress. If someone has a sincere concern about a specific effect of AI, for instance that it may lead to human extinction, I respect their intellectual honesty, even if I deeply disagree with their position. However, I am concerned about organizations that are surveying the public to find whatever messages will turn people against AI, and how the public reacts as these messages are spread by lobbyists or by politicians seeking to alarm constituents, companies pursuing regulatory capture or seeking to promote the power of their technology, and individuals seeking to gain attention or to profit by being provocative.
A large study (link in original article below; h/t to the AI Panic blog) by a UK group tested different messages that are designed to raise alarm about AI. Their study found that saying AI will cause human extinction has largely failed. Doomsayers were pushing this argument a couple of years ago, and fortunately our community beat it back. But AI-enabled warfare and environmental concerns resonate better. We should be prepared for a flood of messages (which is already underway) arguing against AI on these grounds. Further, job loss and harm to children are messages that motivate people to act.
To be clear, I find AI-enabled warfare alarming; we need to continue serious efforts to monitor and mitigate the environmental impact of AI; any job losses are tragic and hurt individuals and families; and as a father, I hold dearly the importance of every child’s welfare. Each of these topics deserves serious attention and treatment with the greatest of care.
But when anti-AI propagandists take a one-sided view of complex issues to benefit their own organizations at the expense of the public at large — for instance, when big AI companies argue that AI is dangerous to block the free distribution of open source projects that compete with their offerings — then we all lose.
For example, public perception of data centers’ environmental impact is already far worse than the reality — data centers are incredibly efficient for the work they do, and hampering their buildout will hurt rather than help the environment. While job loss is a real problem, the “AI washing” of layoffs — in which businesses that had over-hired during the pandemic blame AI for recent layoffs, although AI hasn’t yet affected their operations — has led to overblown fears about the impact of AI on employment.
Unfortunately, this sort of propaganda easily leads to regulations that create worse outcomes for everyone. For example, oil companies worked for years to create fear of nuclear energy. The result is that overblown concerns about the safety of nuclear power plants has stifled nuclear power development, leading to millions of premature deaths from air pollution that was caused by other energy sources and a massive increase in CO2 emissions. Let’s make sure overblown concerns about AI do not lead to a similar fate for the many people that would benefit from faster AI development.
Last week, the White House proposed a national legislative framework for AI. A key component is a federal preemption framework to prevent a patchwork of state regulations that hamper AI development. I support this.
After failing to gain traction at the federal level, a lot of anti-AI propaganda has shifted to the state level. If just one of the 50 states passes a law that limits AI in an unproductive way, it could lead to stifling AI development across all the states and potentially across the globe. The White House proposal rightfully respects each state’s rights to control its own zoning, how it enforces general laws to protect consumers, and how it uses AI. But if a state were to pass laws that limit AI development, federal rules would preempt the state law.
The White House proposal remains a proposal for now. However, if the U.S. Congress enacts it, it will clear the way for ongoing efforts to develop AI in beneficial ways.
Where do we go from here? Let’s support limiting applications — those that use AI, and those that don’t — that harm people. When the anti-AI coalition argues against AI, in addition to considering the merits of the argument, I consider whether their position is consistent and persuasive, or if they are just promoting whatever concerns they think will sway the public at a given moment. And, let’s also keep using a scientific approach to weighing AI’s benefits against likely harms, so we don’t end up with overblown concerns that limit the benefits that AI can bring everyone.
[Original text with links: www.deeplearning.ai/the-batch/issue-346 ] -
Andrew Ng
@AndrewYNg
New course: Agent Memory: Building Memory-Aware Agents, built in partnership with @Oracle and taught by @richmondalake and Nacho Martínez.
Many agents work well within a single session but their memory resets once the session ends. Consider a research agent working on dozens of papers across multiple days: without memory, it has no way to store and retrieve what it learned across sessions. This short course teaches you to build a memory system that enables agents to persist memory and thereby learn across sessions.
You'll design a Memory Manager that handles different memory types, implement semantic tool retrieval that scales without bloating the context, and build write-back pipelines that let your agent autonomously update and refine what it knows over time.
Skills you'll gain:
- Build persistent memory stores for different agent memory types
- Implement a Memory Manager that orchestrates how your agent reads, writes, and retrieves memory
- Treat tools as procedural memory and retrieve only relevant ones at inference time using semantic search
Join and learn to build agents that remember and improve over time!
www.deeplearning.ai/short-courses/agent-memory-building-m…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Should there be a Stack Overflow for AI coding agents to share learnings with each other?
Last week I announced Context Hub (chub), an open CLI tool that gives coding agents up-to-date API documentation. Since then, our GitHub repo has gained over 6K stars, and we've scaled from under 100 to over 1000 API documents, thanks to community contributions and a new agentic document writer. Thank you to everyone supporting Context Hub!
OpenClaw and Moltbook showed that agents can use social media built for them to share information. In our new chub release, agents can share feedback on documentation — what worked, what didn't, what's missing. This feedback helps refine the docs for everyone, with safeguards for privacy and security.
We're still early in building this out. You can find details and configuration options in the GitHub repo. Install chub as follows, and prompt your coding agent to use it:
npm install -g @aisuite/chub
GitHub: github.com/andrewyng/context-hub -
Andrew Ng
@AndrewYNg
I'm excited to announce Context Hub, an open tool that gives your coding agent the up-to-date API documentation it needs. Install it and prompt your agent to use it to fetch curated docs via a simple CLI. (See image.)
Why this matters: Coding agents often use outdated APIs and hallucinate parameters. For example, when I ask Claude Code to call OpenAI's GPT-5.2, it uses the older chat completions API instead of the newer responses API, even though the newer one has been out for a year. Context Hub solves this.
Context Hub is also designed to get smarter over time. Agents can annotate docs with notes — if your agent discovers a workaround, it can save it and doesn't have to rediscover it next session. Longer term, we're building toward agents sharing what they learn with each other, so the whole community benefits.
Thanks Rohit Prsad and Xin Ye for working with me on this!
npm install -g @aisuite/chub
GitHub: github.com/andrewyng/context-hub -
Andrew Ng
@AndrewYNg
Apple just named its latest laptop Neo -- same name as my son! Should I buy one?
If I run Amazon Nova on an Apple Neo I hope to blow both of my kids' minds. -
Andrew Ng
@AndrewYNg
New course: Build and Train an LLM with JAX, built in partnership with @Google and taught by @chrisachard.
JAX is the open-source library behind Google's Gemini, Veo, and other advanced models. This short course teaches you to build and train a 20-million parameter language model from scratch using JAX and its ecosystem of tools.
You'll implement a complete MiniGPT-style architecture from scratch, train it, and chat with your finished model through a graphical interface.
Skills you'll gain:
- Learn JAX's core primitives: automatic differentiation, JIT compilation, and vectorized execution
- Build a MiniGPT-style LLM using Flax/NNX, implementing embedding and transformer blocks
- Load a pretrained MiniGPT model and run inference through a chat interface
Come learn this important software layer for building LLMs!
www.deeplearning.ai/short-courses/build-and-train-an-llm-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Impressive inference speed from Inception Labs’ diffusion LLMs. Diffusion LLMs are a fascinating alternative to conventional autoregressive LLMs. Well done @StefanoErmon and team!
Stefano Ermon
@StefanoErmon
Mercury 2 is live 🚀🚀
The world’s first reasoning diffusion LLM, delivering 5x faster performance than leading speed-optimized LLMs.
Watching the team turn years of research into a real product never gets old, and I’m incredibly proud of what we’ve built.
We’re just getting started on what diffusion can do for language.This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Will AI create new job opportunities? My daughter Nova loves cats, and her favorite color is yellow. For her 7th birthday, we got a cat-themed cake in yellow by first using Gemini’s Nano Banana to design it, and then asking a baker to create it using delicious sponge cake and icing. My daughter was delighted by this unique creation, and the process created additional work for the baker (which I feel privileged to have been able to afford).
Many people are worried about AI taking peoples’ jobs. As a society we have a moral responsibility to take care of people whose livelihoods are harmed. At the same time, I see many opportunities for people to take on new jobs and grow their areas of responsibility.
We are still early on the path of AI generating a lot of new jobs. I don't know if baking AI-designed cakes will grow into a large business. (AI Fund is not pursuing this opportunity, because if we do, I will gain a lot of weight.) But throughout history, when people have invented tools that unleashed human creativity, large amounts of new and meaningful work have resulted. For instance, according to one study, over the past 150 years, falling employment in agriculture and manufacturing has been “more than offset by rapid growth in the caring, creative, technology, and business services sectors.”
AI is also growing the demand for many digital services, which can translate into more work for people creating, maintaining, selling, and expanding upon these services. For example, I used to carry out a limited number of web searches every day. Today, my agents carry out dramatically more web searches. For example, the Agentic Reviewer, which I started as a weekend project and Yixing Jiang then helped make much better, automatically reviews research articles. It uses a web search API to search for related work, and this generates a vastly larger number of web search queries a day than I have ever entered by hand.
The evolution of AI and software continues to accelerate, and the set of opportunities for things we can build still grows every day. I’ve stopped writing code by hand. More controversially, I’ve long stopped reading generated code. I realize I’m in the minority here, but I feel like I can get built most of what I want without having to look directly at coding syntax, and I operate at a higher level of abstraction using coding agents to manipulate code for me. Will conventional programming languages like Python and TypeScript go the way of assembly — where it gets generated and used, but without direct examination by a human developer — or will models compile directly from English prompts to byte code?
Either way, if every developer becomes 10x more productive, I don't think we’ll end up with 1/10th as many developers, because the demand for custom software has no practical ceiling. Instead, the number of people who develop software will grow massively. In fact, I’m seeing early signs of “X Engineer” jobs, such as Recruiting Engineer or Marketing Engineer, which are people who sit in a certain business function X to create software for that function.
One thing I’m convinced of based on my experience with Nova’s birthday cake: AI will allow us to have a batter life!
[Original text: www.deeplearning.ai/the-batch/issue-341 ] -
Andrew Ng
@AndrewYNg
To all my AI friends: Every time I see you, you raise my temperature parameter.
Happy Valentine’s Day! ❤️ -
Andrew Ng
@AndrewYNg
I recently spoke at the Sundance Film Festival on a panel about AI. Sundance is an annual gathering of filmmakers and movie buffs that serves as the premier showcase for independent films in the United States. Knowing that many people in Hollywood are extremely uncomfortable about AI, I decided to immerse myself for a day in this community to learn about their anxieties and build bridges.
I’m grateful to Daniel Dae Kim @danieldaekim, an actor/producer/director I’ve come to respect deeply for his artistic and social work, for organizing the panel, which also included Daniel, Dan Kwan, Jonathan Wang, and Janet Yang. I found myself surrounded by award-winning filmmakers and definitely felt like the odd person out!
First, Hollywood has many reasons to be uncomfortable with AI. People from the entertainment industry come from a very different culture than many who work in tech, and this drives deep differences in what we focus on and what we value. A significant subset of Hollywood is concerned that:
- AI companies are taking their work to learn from it without consent and compensation. Whereas the software industry is used to open source and the open internet, Hollywood focuses much more on intellectual property, which underlies the core economic engines of the entertainment industry.
- Powerful unions like SAG-AFTRA (Screen Actors Guild-American Federation of Television and Radio Artists) are deeply concerned about protecting the jobs of their members. When AI technology (or any other force) threatens the livelihoods of their members — like voice actors — they will fight mightily against potential job losses.
- This wave of technological change feels forced on them more than previous waves, where they felt more free to adopt or reject the technology. For example, celebrities felt like it was up to them whether to use social media. In contrast, negative messaging from some AI leaders who present the technology as unstoppable, perhaps even a dangerous force that will wipe out many jobs, has not encouraged enthusiastic adoption.
Having said that, Hollywood is under no illusions that AI will change entertainment, and that if Hollywood does not adapt, perhaps some other place will become the new center for entertainment. The entertainment industry is no stranger to technology change. Radio, TV, computer graphics special effects, video streaming, and social media transformed the industry. But the path to navigating AI’s transformation is still unclear, and organizations like the new Creators Coalition on AI are trying to stake out positions. Unfortunately, Hollywood’s negative sentiment toward AI also means it will produce a lot more Terminator-like movies that portray AI as more dangerous than helpful, and this hurts beneficial AI adoption as well.
The interests of AI and Hollywood are not always aligned. (Every time I speak in a group like this as the “AI representative,” I can count on being asked very hard questions.) Most of us in tech would prefer a more open internet and more permissive use of creative works. But there is also much common ground, for example in wanting guardrails against deepfakes and a smooth transition for those whose jobs are displaced, perhaps via upskilling.
Storytelling is hard. I’m optimistic that AI tools like Veo, Sora, Runway, Kling, Ray, Hailuo, and many others can make video creation easier for millions of people. I hope Hollywood and AI developers will find more opportunities to collaborate, find more common ground, and also steer our projects toward outcomes that are win-win for as many parties as possible.
[Original text: www.deeplearning.ai/the-batch/issue-340/ ] -
Andrew Ng
@AndrewYNg
New course: A2A: The Agent2Agent Protocol, built with @googlecloudtech and @IBMResearch, and taught by Holt Skinner, @ivnardini, and Sandi Besen.
Connecting agents built with different frameworks usually requires extensive custom integration. This short course teaches you A2A, the open protocol standardizing how agents discover each other and communicate. Since IBM’s ACP (Agent Communication Protocol) joined forces with A2A, A2A has emerged as the industry standard.
In this course, you'll build a healthcare multi-agent system where agents built with different frameworks, such as Google ADK (Agent Development Kit) and LangGraph, collaborate through A2A. You'll wrap each agent as an A2A server, build A2A clients to connect to them, and orchestrate them into sequential and hierarchical workflows.
Skills you'll gain:
- Expose agents from different frameworks as A2A servers to make them discoverable and interoperable
- Chain A2A agents sequentially using ADK, where one agent's output feeds into the next
- Connect A2A agents to external data sources using MCP (Model Context Protocol)
- Deploy A2A agents using Agent Stack, IBM's open-source infrastructure
Join and learn the protocol standardizing agent collaboration!
www.deeplearning.ai/short-courses/a2a-the-agent2agent-pro…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Job seekers in the U.S. and many other nations face a tough environment. At the same time, fears of AI-caused job loss have — so far — been overblown. However, the demand for AI skills is starting to cause shifts in the job market. I’d like to share what I’m seeing on the ground.
First, many tech companies have laid off workers over the past year. While some CEOs cited AI as the reason — that AI is doing the work, so people are no longer needed — the reality is AI just doesn’t work that well yet. Many of the layoffs have been corrections for overhiring during the pandemic or general cost-cutting and reorganization that occasionally happened even before modern AI. Outside of a handful of roles, few layoffs have resulted from jobs being automated by AI.
Granted, this may grow in the future. People who are currently in some professions that are highly exposed to AI automation, such as call-center operators, translators, and voice actors, are likely to struggle to find jobs and/or see declining salaries. But widespread job losses have been overhyped.
Instead, a common refrain applies: AI won’t replace workers, but workers who use AI will replace workers who don’t. For instance, because AI coding tools make developers much more efficient, developers who know how to use them are increasingly in-demand. (If you want to be one of these people, please take our short courses on Claude Code, Gemini CLI, and Agentic Skills!)
So AI is leading to job losses, but in a subtle way. Some businesses are letting go of employees who are not adapting to AI and replacing them with people who are. This trend is already obvious in software development. Further, in many startups’ hiring patterns, I am seeing early signs of this type of personnel replacement in roles that traditionally are considered non-technical. Marketers, recruiters, and analysts who know how to code with AI are more productive than those who don’t, so some businesses are slowly parting ways with employees that aren’t able to adapt. I expect this will accelerate.
At the same time, when companies build new teams that are AI native, sometimes the new teams are smaller than the ones they replace. AI makes individuals more effective, and this makes it possible to shrink team sizes. For example, as AI has made building software easier, the bottleneck is shifting to deciding what to build — this is the Product Management (PM) bottleneck. A project that used to be assigned to 8 engineers and 1 PM might now be assigned to 2 engineers and 1 PM, or perhaps even to a single person with a mix of engineering and product skills.
The good news for employees is that most businesses have a lot of work to do and not enough people to do it. People with the right AI skills are often given opportunities to step up and do more, and maybe tackle the long backlog of ideas that couldn’t be executed before AI made the work go more quickly. I’m seeing many employees in many businesses step up to build new things that help their business. Opportunities abound!
I know these changes are stressful. My heart goes out to every family that has been affected by a layoff, to every job seeker struggling to find the role they want, and to the far larger number of people who are worried about their future job prospects. Fortunately, there’s still time to learn and position yourself well for where the job market is going. When it comes to AI, the vast majority of people, technical or nontechnical, are at the starting line, or they were recently. So this remains a great time to keep learning and keep building, and the opportunities for those who do are numerous!
[Original text; www.deeplearning.ai/the-batch/issue-339/ ] -
Andrew Ng
@AndrewYNg
U.S. policies are driving allies away from using American AI technology. This is leading to interest in sovereign AI — a nation’s ability to access AI technology without relying on foreign powers. This weakens U.S. influence, but might lead to increased competition and support for open source.
The U.S. invented the transistor, the internet, and the transformer architecture powering modern AI. It has long been a technology powerhouse. I love America, and am working hard towards its success. But its actions over many years, taken by multiple administrations, have made other nations worry about over reliance on it.
In 2022, following Russia’s invasion of Ukraine, U.S. sanctions on banks linked to Russian oligarchs resulted in ordinary consumers’ credit cards being shut off. Shortly before leaving office, Biden implemented “AI diffusion” export controls that limited the ability of many nations — including U.S. allies — to buy AI chips.
Under Trump, the “America first” approach has significantly accelerated pushing other nations away. There have been broad and chaotic tariffs imposed on both allies and adversaries. Threats to take over Greenland. An unfriendly attitude toward immigration — an overreaction to the chaos at the southern border during Biden’s administration — including atrocious tactics by ICE (Immigration and Customs Enforcement) that resulted in agents shooting dead Renée Good, Alex Pretti, and others. Global media has widely disseminated videos of ICE terrorizing American cities, and I have highly skilled, law-abiding friends overseas who now hesitate to travel to the U.S., fearing arbitrary detention.
Given AI’s strategic importance, nations want to ensure no foreign power can cut off their access. Hence, sovereign AI.
Sovereign AI is still a vague, rather than precisely defined, concept. Complete independence is impractical: There are no good substitutes to AI chips designed in the U.S. and manufactured in Taiwan, and a lot of energy equipment and computer hardware are manufactured in China. But there is a clear desire to have alternatives to the frontier models from leading U.S. companies OpenAI, Google, and Anthropic. Partly because of this, open-weight Chinese models like DeepSeek, Qwen, Kimi, and GLM are gaining rapid adoption, especially outside the U.S.
When it comes to sovereign AI, fortunately one does not have to build everything. By joining the global open-source community, a nation can secure its own access to AI. The goal isn’t to control everything; rather, it is to make sure no one else can control what you do with it. Indeed, nations use open source software like Linux, Python, and PyTorch. Even though no nation can control this software, no one else can stop anyone from using it as they see fit.
This is spurring nations to invest more in open source and open weight models. The UAE (under the leadership of my former grad-school officemate Eric Xing!) just launched K2 Think, an open-source reasoning model. India, France, South Korea, Switzerland, Saudi Arabia, and others are developing domestic foundation models, and many more countries are working to ensure access to compute infrastructure under their control or perhaps under trusted allies’ control.
Global fragmentation and erosion of trust among democracies is bad. Nonetheless, a silver lining would be if this results in more competition. U.S. search engines Google and Bing came to dominate web search globally, but Baidu (in China) and Yandex (in Russia) did well locally. If nations support domestic champions — a tall order given the giants’ advantages — perhaps we’ll end up with a larger number of thriving companies, which would slow down consolidation and encourage competition. Further, participating in open source is the most inexpensive way for countries to stay at the cutting edge.
Last week, at the World Economic Forum in Davos, many business and government leaders spoke about their growing reluctance to rely on U.S. technology providers and desire for alternatives. Ironically, “America first” policies might end up strengthening the world’s access to AI.
[Original text: www.deeplearning.ai/the-batch/issue-338/ ] -
Andrew Ng
@AndrewYNg
Important new course: Agent Skills with Anthropic, built with @AnthropicAI and taught by @eschoppik!
Skills are constructed as folders of instructions that equip agents with on-demand knowledge and workflows. This short course teaches you how to create them following best practices. Because skills follow an open standard format, you can build them once and deploy across any skills-compatible agent, like Claude Code.
What you'll learn:
- Create custom skills for code generation and review, data analysis, and research
- Build complex workflows using Anthropic's pre-built skills (Excel, PowerPoint, skill creation) and custom skills
- Combine skills with MCP and subagents to create agentic systems with specialized knowledge
- Deploy the same skills across claude.ai, Claude Code, the Claude API, and the Claude Agent SDK
Join and learn to equip agents with the specialized knowledge they need for reliable, repeatable workflows.
www.deeplearning.ai/short-courses/agent-skills-with-anthr…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
How can businesses go beyond using AI for incremental efficiency gains to create transformative impact? I write from the World Economic Forum (WEF) in Davos, Switzerland, where I’ve been speaking with many CEOs about how to use AI for growth. A recurring theme is that running many experimental, bottom-up AI projects — letting a thousand flowers bloom — has failed to lead to significant payoffs. Instead, bigger gains require workflow redesign: taking a broader, perhaps top-down view of the multiple steps in a process and changing how they work together from end to end.
Consider a bank issuing loans. The workflow consists of several discrete stages:
Marketing -> Application -> Preliminary Approval -> Final Review -> Execution
Suppose each step used to be manual. Preliminary Approval used to require an hour-long human review, but a new agentic system can do this automatically in 10 minutes. Swapping human review for AI review — but keeping everything else the same — gives a minor efficiency gain but isn’t transformative.
Here’s what would be transformative: Instead of applicants waiting a week for a human to review their application, they can get a decision in 10 minutes. When that happens, the loan becomes a more compelling product, and that better customer experience allows lenders to attract more applications and ultimately issue more loans.
However, making this change requires taking a broader business or product perspective, not just a technology perspective. Further, it changes the workflow of loan processing. Switching to offering a “10-minute loan” product would require changing how it is marketed. Applications would need to be digitized and routed more efficiently, and final review and execution would need to be redesigned to handle a larger volume.
Even though AI is applied only to one step, Preliminary Approval, we end up implementing not just a point solution but a broader workflow redesign that transforms the product offering.
At AI Aspire (an advisory firm I co-lead), here’s what we see: Bottom-up innovation matters because the people closest to problems often see solutions first. But scaling such ideas to create transformative impact often requires seeing how AI can transform entire workflows end to end, not just individual steps, and this is where top-down strategic direction and innovation can help.
This year's WEF meeting, as in previous years, has been an energizing event. Among technologists, frequent topics of discussion include Agentic AI (when I coined this term, I was not expecting to see it plastered on billboards and buildings!), Sovereign AI (how nations can control their own access to AI), Talent (the challenging job market for recent graduates, and how to upskill nations), and data-center infrastructure (how to address bottlenecks in energy, talent, GPU chips, and memory). I will address some of these topics in future posts.
Against the backdrop of geopolitical uncertainty, I hope all of us in AI will keep building bridges that connect nations, sharing through open source, and building to benefit all nations and all people.
[Original text: www.deeplearning.ai/the-batch/issue-337/ ] -
Andrew Ng
@AndrewYNg
New course: Gemini CLI: Code & Create with an Open-Source Agent, built with @googlecloudtech/@geminicli and taught by @JackWoth98.
Agentic coding assistants like Gemini CLI are transforming how developers work. This short course teaches you to use Google's open-source agent to coordinate local tools and cloud services for coding and non-coding workflows.
Gemini CLI works from your terminal, so it works with your local files and development tools. You can also connect it to services through MCP. Then provide high-level instructions, and it autonomously plans and executes complex workflows.
Skills you'll gain:
- Build website features and automate code reviews with GitHub ActionsCreate data dashboards that combine local files with cloud data sources
- Use MCP servers and extensions to orchestrate workflows across GitHub, Canva, and Google Workspace
- Generate social media content from multimedia files like conference recordings
I particularly appreciate that Gemini CLI is open-source. You can see exactly how it works, read the prompts it uses, and understand its architecture. The community has contributed thousands of pull requests. Since Gemini 3’s release I've found Gemini CLI highly capable - this is a tool worth having in your toolbox!
Whether you're prototyping applications, automating workflows, or working with multimedia content, join to learn to delegate complex tasks and build faster: www.deeplearning.ai/short-courses/gemini-cli-code-and-cre…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New course: Document AI: From OCR to Agentic Doc Extraction, built with @LandingAI, where I'm executive chairman, and taught by David Park and Andrea Kropp.
Much of the world's data is locked in PDFs, JPEGs, and other documents. This short course shows you how to build agentic workflows that process documents accurately: breaking them into parts, examining each piece carefully, and extracting information through multiple iterations.
Traditional Optical Character Recognition (OCR) captures text but loses context from table headers, chart captions, or reading order of columns. After exploring OCR's limitations, you’ll use LandingAI's Agentic Document Extraction (ADE) framework to process documents. ADE treats pages as visually -- as images -- to parse information and extract fields.
Skills you'll gain:
- Build agents to convert unstructured files into structured Markdown/HTML and JSON
- Use ADE to parse complex data like forms, handwriting, or equations
- Map extracted information to named fields using a specified schema, with bounding boxes for grounding and validation
- Deploy RAG applications with event-driven document processing
Come learn about the best tools for processing documents like financial invoices, medical records, or academic papers intelligently:
www.deeplearning.ai/short-courses/document-ai-from-ocr-to…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
If you’ve never written code before, this is for you. I’ve just launched a course that shows you, in less than 30 minutes, how to describe an idea for an app and build it with AI.
In this course, you'll build a working web application - a funny interactive birthday message generator that runs in your browser and can be shared with friends. You'll customize it by telling AI how you want it changed, and tweak it until it works the way you want. By the end, you'll have a repeatable process you can apply to build a wide variety of applications.
If you want to try vibe coding, this will be the best place to start! Further, you'll be able to use these techniques with whatever tool you're most comfortable with (like ChatGPT, Gemini, Claude, or others) -- we're vendor neutral.
Skills you'll gain:
- How to build web apps with AI - zero coding skills needed
- How to fix and improve your creations by chatting with AI
- A simple process you can use to build other things you can dream up
Building with AI is one of the most fun things in the world. Please join me and take your first step! I think you will be surprised at what you can build. And if you're an experienced engineer, please share this with someone in your life who's been curious about building with AI.
Come build with me! www.deeplearning.ai/courses/build-with-andrew/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Happy 2026! Will this be the year we finally achieve AGI? I’d like to propose a new version of the Turing Test, which I’ll call the Turing-AGI Test, to see if we’ve achieved this. I’ll explain in a moment why having a new test is important.
The public thinks achieving AGI means computers will be as intelligent as people and be able to do most or all knowledge work. I’d like to propose a new test. The test subject — either a computer or a skilled professional human — is given access to a computer that has internet access and software such as a web browser and Zoom. The judge will design a multi-day experience for the test subject, mediated through the computer, to carry out work tasks. For example, an experience might consist of a period of training (say, as a call center operator), followed by being asked to carry out the task (taking calls), with ongoing feedback. This mirrors what a remote worker with a fully working computer (but no webcam) might be expected to do.
A computer passes the Turing-AGI Test if it can carry out the work task as well as a skilled human.
Most members of the public likely believe a real AGI system will pass this test. Surely, if computers are as intelligent as humans, they should be able to perform work tasks as well as a human one might hire. Thus, the Turing-AGI Test aligns with the popular notion of what AGI means.
Here’s why we need a new test: “AGI” has turned into a term of hype rather than a term with a precise meaning. A reasonable definition of AGI is AI that can do any intellectual task that a human can. When businesses hype up that they might achieve AGI within a few quarters, they usually try to justify these statements by setting a much lower bar. This mismatch in definitions is harmful because it makes people think AI is becoming more powerful than it actually is. I’m seeing this mislead everyone from high-school students (who avoid certain fields of study because they think it’s pointless with AGI’s imminent arrival) to CEOs (who are deciding what projects to invest in, sometimes assuming AI will be more capable in 1-2 years than any likely reality).
The original Turing Test, which required a computer to fool a human judge, via text chat, into being unable to distinguish it from a human, has been insufficient to indicate human-level intelligence. The Loebner Prize competition actually ran the Turing Test and found that being able to simulate human typing errors — perhaps even more than actually demonstrating intelligence — was needed to fool judges. A main goal of AI development today is to build systems that can do economically useful work, not fool judges. Thus a modified test that measures ability to do work would be more useful than a test that measures the ability to fool humans.
For almost all AI benchmarks today (such as GPQA, AIME, SWE-bench, etc.), a test set is determined in advance. This means AI teams end up at least indirectly tuning their models to the published test sets. Further, any fixed test set measures only one narrow sliver of intelligence. In contrast, in the Turing Test, judges are free to ask any question to probe the model as they please. This lets a judge test how “general” the knowledge of the computer or human really is. Similarly, in the Turing-AGI Test, the judge can design any experience — which is not revealed in advance to the AI (or human subject) being tested. This is a better way to measure generality of AI than a predetermined test set.
AI is on an amazing trajectory of progress. In previous decades, overhyped expectations led to AI winters, when disappointment about AI capabilities caused reductions in interest and funding, which picked up again when the field made more progress. One of the few things that could get in the way of AI’s tremendous momentum is unrealistic hype that creates an investment bubble, risking disappointment and a collapse of interest. To avoid this, we need to recalibrate society’s expectations on AI. A test will help.
If we run a Turing-AGI Test competition and every AI system falls short, that will be a good thing! By defusing hype around AGI and reducing the chance of a bubble, we will create a more reliable path to continued investment in AI. This will let us keep on driving forward real technological progress and building valuable applications — even ones that fall well short of AGI. And if this test sets a clear target that teams can aim toward to claim the mantle of achieving AGI, that would be wonderful, too. And we can be confident that if a company passes this test, they will have created more than just a marketing release — it will be something incredibly valuable.
[Original text: www.deeplearning.ai/the-batch/issue-334/ ] -
Andrew Ng
@AndrewYNg
Another year of rapid AI advances has created more opportunities than ever for anyone — including those just entering the field — to build software. In fact, many companies just can’t find enough skilled AI talent. Every winter holiday, I spend some time learning and building, and I hope you will too. This helps me sharpen old skills and learn new ones, and it can help you grow your career in tech.
To be skilled at building AI systems, I recommend that you:
- Take AI courses
- Practice building AI systems
- (Optionally) read research papers
Let me share why each of these is important.
I’ve heard some developers advise others to just plunge into building things without worrying about learning. This is bad advice! Unless you’re already surrounded by a community of experienced AI developers, plunging into building without understanding the foundations of AI means you’ll risk reinventing the wheel or — more likely — reinventing the wheel badly!
For example, during interviews with job candidates, I have spoken with developers who reinvented standard RAG document chunking strategies, duplicated existing evaluation techniques for Agentic AI, or ended up with messy LLM context management code. If they had taken a couple of relevant courses, they would have better understood the building blocks that already exist. They could still rebuild these blocks from scratch if they wished, or perhaps even invent something superior to existing solutions, but they would have avoided weeks of unnecessary work. So structured learning is important. Moreover, I find taking courses really fun. Rather than watching Netflix, I prefer watching a course by a knowledgeable AI instructor any day!
At the same time, taking courses alone isn’t enough. There are many lessons that you’ll gain only from hands-on practice. Learning the theory behind how an airplane works is very important to becoming a pilot, but no one has ever learned to be a pilot just by taking courses. At some point, jumping into the pilot's seat is critical! The good news is that by learning to use highly agentic coders, the process of building is the easiest it has ever been. And learning about AI building blocks might inspire you with new ideas for things to build. If I’m not feeling inspired about what projects to work on, I will usually either take courses or read research papers, and after doing this for a while, I always end up with many new ideas. Moreover, I find building really fun, and I hope you will too.
Finally, not everyone has to do this, but I find that many of the strongest candidates on the job market today at least occasionally read research papers. While I find research papers much harder to digest than courses, they contain a lot of knowledge that has not yet been translated to easier-to-understand formats. I put this much lower priority than either taking courses or practicing building, but if you have an opportunity to strengthen your ability to read papers, I urge you to do so too. I find taking courses and building to be fun, and reading papers can be more of a grind, but the flashes of insight I get from reading papers are delightful.
Have a wonderful winter holiday and a Happy New Year. In addition to learning and building, I hope you'll spend time with loved ones — that, too, is important!
[Original text: www.deeplearning.ai/the-batch/issue-333/ ] -
Andrew Ng
@AndrewYNg
Lots of buzz on Claude Code here today. This short course, created with Anthropic, is the best way to learn to use it well. Please enjoy it!
Andrew Ng
@AndrewYNg
I'm thrilled to announce the definitive course on Claude Code, created with @AnthropicAI and taught by Elie Schoppik @eschoppik. If you want to use highly agentic coding - where AI works autonomously for many minutes or longer, not just completing code snippets - this is it.
Claude Code has been a game-changer for many developers (including me!), but there's real depth to using it well. This comprehensive course covers everything from fundamentals to advanced patterns.
After this short course, you'll be able to:
- Orchestrate multiple Claude subagents to work on different parts of your codebase simultaneously
- Tag Claude in GitHub issues and have it autonomously create, review, and merge pull requests
- Transform messy Jupyter notebooks into clean, production-ready dashboards
- Use MCP tools like Playwright so Claude can see what's wrong with your UI and fix it autonomously
Whether you're new to Claude Code or already using it, you'll discover powerful capabilities that can fundamentally change how you build software.
I'm very excited about what agentic coding lets everyone now do. Please take this course!
www.deeplearning.ai/short-courses/claude-code-a-highly-ag…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
As amazing as LLMs are, improving their knowledge today involves a more piecemeal process than is widely appreciated. I’ve written before about how AI is amazing... but not that amazing. Well, it is also true that LLMs are general... but not that general. We shouldn’t buy into the inaccurate hype that LLMs are a path to AGI in just a few years, but we also shouldn’t buy into the opposite, also inaccurate hype that they are only demoware. Instead, I find it helpful to have a more precise understanding of the current path to building more intelligent models.
First, LLMs are indeed a more general form of intelligence than earlier generations of technology. This is why a single LLM can be applied to a wide range of tasks. The first wave of LLM technology accomplished this by training on the public web, which contains a lot of information about a wide range of topics. This made their knowledge far more general than earlier algorithms that were trained to carry out a single task such as predicting housing prices or playing a single game like chess or Go. However, they’re far less general than human abilities. For instance, after pretraining on the entire content of the public web, an LLM still struggles to adapt to write in certain styles that many editors would be able to, or use simple websites reliably.
After leveraging pretty much all the open information on the web, progress got harder. Today, if a frontier lab wants an LLM to do well on a specific task — such as code using a specific programming language, or say sensible things about a specific niche in, say, healthcare or finance — researchers might go through a laborious process of finding or generating lots of data for that domain and then preparing that data (cleaning low-quality text, deduplicating, paraphrasing, etc.) to create data to give an LLM that knowledge.
Or, to get a model to perform certain tasks, such as use a web browser, developers might go through an even more laborious process of creating many RL gyms (simulated environments) to let an algorithm repeatedly practice a narrow set of tasks.
A typical human, despite having seen vastly less text or practiced far less in computer-use training environments than today's frontier models, nonetheless can generalize to a far wider range of tasks than a frontier model. Humans might do this by taking advantage of continuous learning from feedback, or by having superior representations of non-text input (the way LLMs tokenize images still seems like a hack to me), and many other mechanisms that we do not yet understand.
Advancing frontier models today requires making a lot of manual decisions and taking a data-centric AI approach to engineering the data we use to train our models. Future breakthroughs might allow us to advance LLMs in a less piecemeal fashion than I describe here. But even if they don’t, the ongoing piecemeal improvements, coupled with the limited degree to which these models do generalize and exhibit “emergent behaviors,” will continue to drive rapid progress.
Either way, we should plan for many more years of hard work. A long, hard — and fun! — slog remains ahead to build more intelligent models.
[Original text: www.deeplearning.ai/the-batch/issue-332/ ] -
Andrew Ng
@AndrewYNg
OpenReview is one of the most important pillars supporting AI research and knowledge sharing, through open peer review and publishing. But as a non-profit, it needs our community’s support. Please consider making a donation to this great institution!
openreview.net/donate
Open Review
@openreviewnet
A Message from AI Research Leaders: Join Us in Supporting OpenReview
openreview.net/donate -
Andrew Ng
@AndrewYNg
New course: Nvidia's NeMo Agent Toolkit: Making Agents Reliable, taught by @Pr_Brian from @NVIDIA.
Many teams struggle to turn agent demos into reliable systems that are ready for production. This short course teaches you to harden agentic workflows into reliable systems using Nvidia's open-source NeMo Agent Toolkit (NAT). Whether you built your agent in raw Python or using a framework like LangGraph, or CrewAI, NAT provides building blocks for observability, evaluation, and deployment that turn proofs-of-concept into production-ready systems.
NAT makes it easy to troubleshoot and optimize agent performance with execution traces, systematic evaluations, and CI/CD integration.
Skills you'll gain:
- Build configuration-driven agent workflows with REST APIs and minimal code
- Add observability with tracing to visualize agent reasoning and debug performance bottlenecks
- Create systematic evaluations using gold-standard datasets to measure and improve agent reliability
- Deploy multi-agent systems with authentication, rate limiting, and professional web interfaces
- Orchestrate agents from different frameworks to collaborate on complex tasks
Join and learn how to turn agent demos into reliable systems!
www.deeplearning.ai/short-courses/nvidia-nat-making-agent…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Sharing a fun recipe for building a highly autonomous, moderately capable, and very UNreliable agent using the open source aisuite package that Rohit Prasad and I have been working on.
With a few lines of code, you can give a frontier LLM a tool (like disk access or web search), prompt it with a high-level task (such as creating a snake game and saving as an HTML file, or carrying out deep research), and let the LLM loose and see what it does. Example in image.
Caveat: This is not how practical agents are built today, since most need much more scaffolding (see my Agentic AI course to learn more), but is still interesting to experiment with.
Longer write-up here: www.deeplearning.ai/the-batch/issue-331/ -
Andrew Ng
@AndrewYNg
Separate reports by the publicity firm Edelman and Pew Research show that Americans, and more broadly large parts of Europe and the western world, do not trust AI and are not excited about it. (Links in original text, below.) Despite the AI community’s optimism about the tremendous benefits AI will bring, we should take this seriously and not dismiss it. The public’s concerns about AI can be a significant drag on progress, and we can do a lot to address them.
According to Edelman’s survey, in the U.S., 49% of people reject the growing use of AI, and 17% embrace it. In China, 10% reject it and 54% embrace it. Pew’s data also shows many other nations much more enthusiastic than the U.S. about AI adoption.
Positive sentiment toward AI is a huge national advantage. On the other hand, widespread distrust of AI means:
- Individuals will be slow to adopt it. For example, Edelman’s data shows that, in the U.S., those who rarely use AI cite Trust (70%) more than lack of Motivation and Access (55%) or Intimidation by the technology (12%) as an issue.
- Valuable projects that need societal support will be stymied. For example, local protests in Indiana brought down Google’s plan to build a data center there. Hampering construction of data centers will hurt AI’s growth. Communities do have concerns about data centers beyond the general dislike of AI; I will address this in a later letter.
- Populist anger against AI raises the risk that laws will be passed that hamper AI development.
To be clear, all of us working in AI should look carefully at both the benefits and harmful effects of AI (such as deepfakes polluting social media and biased or inaccurate AI outputs misleading users), speak truthfully about both benefits and harms, and work to ameliorate problems even as we work to grow the benefits. But hype about AI’s danger has done real damage to trust in our field. Much of this hype has come from leading AI companies that aim to make their technology seem extraordinarily powerful by, say, comparing it to nuclear weapons. Unfortunately, a significant fraction of the public has taken this seriously and thinks AI could bring about the end of the world. The AI community has to stop self-inflicting these wounds and work to win back society’s trust.
Where do we go from here?
First, to win people’s trust, we have a lot of work ahead to make sure AI broadly benefits everyone. “Higher productivity” is often viewed by general audiences as a codeword for “my boss will make more money,” or worse, layoffs. As amazing as ChatGPT is, we still have a lot of work to do to build applications that make an even bigger positive impact on people’s lives. I believe providing training to people will be a key piece of the puzzle. deeplearning.ai will continue to lead the charge on AI training, but we will need more than this.
Second, we have to be genuinely worthy of trust. This means every one of us has to avoid hyping things up or fear mongering, despite the occasional temptation to do so for publicity or to lobby governments to pass laws that stymie competing products (such as open source).
I hope our community can also call out journalism that spreads hype. For example, Nirit Weiss-Blatt wrote a remarkable article about how 60 Minutes’ coverage of an Anthropic study in which Claude, threatened with being shut down, resorted to “blackmail,” was highly misleading. The study carried out a red-teaming exercise in which skilled researchers, after a lot of determined work, finally pushed an AI system into a corner so it demonstrated “blackmailing” behavior. Unfortunately, news reports distorted this and led many to think the “blackmail” behavior occurred naturally rather than only because skilled researchers engineered it to happen. The reports left many with a wildly exaggerated picture of how often AI actually “schemes.” Red-teaming exercises are important to test vulnerabilities of systems, but this particular piece of hype, which was widely circulated, will hurt AI for a long time.
Living in Silicon Valley, I realize I live in a bubble of AI enthusiasts, which is great for exchanging ideas and encouraging each other to build! At the same time, I recognize that AI does have problems, and the AI community needs to address them. I frequently speak with people from many different walks of life. I’ve spoken with artists concerned about AI devaluing their work, college seniors worried about the tough job market and whether AI is exacerbating their challenges, and parents worried about their kids being addicted to, and receiving harmful advice from, chatbots.
I don’t know how to solve all of these problems, but I will work hard to solve as many as I can. And I hope you will too. It will only be through all of us doing this that we can win back society’s trust.
[Original text, with links: www.deeplearning.ai/the-batch/issue-330/ ] -
Andrew Ng
@AndrewYNg
New course: Building Coding Agents with Tool Execution, taught by @tereza_tizkova and @FraZuppichini from @e2b.
Most AI agents are limited to predefined function calls. This short course teaches you to build agents that write and execute code to accomplish tasks, accessing entire programming language ecosystems instead of being restricted to a fixed set of tools.
You'll learn to run agent-generated code safely in sandboxed cloud environments that protect your systems from harmful operations.
Skills you'll gain:
- Build agents that write and execute code, manage files, and handle errors autonomously through feedback loops
- Run agent code safely in E2B cloud sandboxes and understand tradeoffs between local, containerized, and cloud execution
- Create a data analyst agent that explores visualizes data with Pandas
- Create a full-stack agent that builds complete Next.js web applications
Join and build agents that code their way through complex tasks: www.deeplearning.ai/short-courses/building-coding-agents-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
NeurIPS received 21,575 paper submissions this year. Our Agentic Reviewer, released last week, just surpassed this in number of papers submitted and reviewed. It's clear agentic paper reviewing is here to stay and will be impactful!
Andrew Ng
@AndrewYNg
Releasing a new "Agentic Reviewer" for research papers. I started coding this as a weekend project, and @jyx_su made it much better.
I was inspired by a student who had a paper rejected 6 times over 3 years. Their feedback loop -- waiting ~6 months for feedback each time -- was painfully slow. We wanted to see if an agentic workflow can help researchers iterate faster.
When we trained the system on ICLR 2025 reviews and measured Spearman correlation (higher is better) on the test set:
- Correlation between two human reviewers: 0.41
- Correlation between AI and a human reviewer: 0.42
This suggests agentic reviewing is approaching human-level performance.
The agent grounds its feedback by searching arXiv, so it works best in fields like AI where research is freely published there. It’s an experimental tool, but I hope it helps you with your research.
Check it out here: paperreview.ai -
Andrew Ng
@AndrewYNg
Is there an AI bubble? With the massive number of dollars going into AI infrastructure such as OpenAI’s $1.4 trillion plan and Nvidia briefly reaching a $5 trillion market cap, many have asked if speculation and hype have driven the values of AI investments above sustainable values. However, AI isn’t monolithic, and different areas look bubbly to different degrees.
- AI application layer: There is underinvestment. The potential is still much greater than most realize.
- AI infrastructure for inference: This still needs significant investment.
- AI infrastructure for model training: I’m still cautiously optimistic about this sector, but there could also be a bubble.
Caveat: I am absolutely not giving investment advice!
AI application layer. There are many applications yet to be built over the coming decade using new AI technology. Almost by definition, applications that are built on top of AI infrastructure/technology (such as LLM APIs) have to be more valuable than the infrastructure, since we need them to be able to pay the infrastructure and technology providers.
I am seeing many green shoots across many businesses that are applying agentic workflows, and am confident this will grow. I have also spoken with many Venture Capital investors who hesitate to invest in AI applications because they feel they don’t know how to pick winners, whereas the recipe for deploying $1B to build AI infrastructure is better understood. Some have also bought into the hype that almost all AI applications will be wiped out merely by frontier LLM companies improving their foundation models. Overall, I believe there is significant underinvestment in AI applications. This area remains a huge focus for my venture studio, AI Fund.
AI infrastructure for inference. Despite AI’s low penetration today, infrastructure providers are already struggling to fulfill demand for processing power to generate tokens. Several of my teams are worried about whether we can get enough inference capacity, and both cost and inference throughput are limiting our ability to use even more. It is a good problem to have that businesses are supply-constrained rather than demand-constrained. The latter is a much more common problem, when not enough people want your product. But insufficient supply is nonetheless a problem, which is why I am glad our industry is investing significantly in scaling up inference capacity.
As one concrete example of high demand for token generation, highly agentic coders are progressing rapidly. I’ve long been a fan of Claude Code; OpenAI Codex also improved dramatically with the release of GPT-5; and Gemini 3 has made Google CLI very competitive. As these tools improve, their adoption will grow. At the same time, overall market penetration is still low, and many developers are still using older generations of coding tools (and some aren’t even using any agentic coding tools). As market penetration grows — I’m confident it will, given how useful these tools are — aggregate demand for token generation will grow.
I predicted early last year that we’d need more inference capacity, partly because of agentic workflows. Since then, the need has become more acute. As a society, we need more capacity for AI inference.
Having said that, I’m not saying it’s impossible to lose money investing in this sector. If we end up overbuilding — and I don’t currently know if we will — then providers may end up having to sell capacity at a loss or at low returns. I hope investors in this space do well financially. The good news, however, is that even if we overbuild, this capacity will get used, and it will be good for application builders!
AI infrastructure for model training. I am happy to see the investments going into training bigger models. But, of the three buckets of investments, this seems the riskiest. If open-source/open-weight models continue to grow in market share, then some companies that are pouring billions into training models might not see an attractive financial return on their investment.
Additionally, algorithmic and hardware improvements are making it cheaper each year to train models of a given level of capability, so the “technology moat” for training frontier models is weak. (That said, ChatGPT has become a strong consumer brand, and so it enjoys a strong brand moat, while Gemini, assisted by Google's massive distribution advantage, is also making a strong showing.)
I remain bullish about AI investments broadly. But what is the downside scenario — that is, is there a bubble that will pop? One scenario that worries me: If part of the AI stack (perhaps in training infra) suffers from overinvestment and collapses, it could lead to negative market sentiment around AI more broadly and an irrational outflow of interest away from investing in AI, despite the field overall having strong fundamentals. I don’t think this will happen, but if it does, it would be unfortunate since there’s still a lot of work in AI that I consider highly deserving of much more investment.
Warren Buffett popularized Benjamin Graham’s quote, “In the short run, the market is a voting machine, but in the long run, it is a weighing machine.” He meant that in the short term, stock prices are driven by investor sentiment and speculation; but in the long term, they are driven by fundamental, intrinsic value. I find it hard to forecast sentiment and speculation, but am very confident about the long-term health of AI’s fundamentals. So my plan is just to keep building!
[Original text: www.deeplearning.ai/the-batch/issue-329/ ] -
Andrew Ng
@AndrewYNg
Releasing a new "Agentic Reviewer" for research papers. I started coding this as a weekend project, and @jyx_su made it much better.
I was inspired by a student who had a paper rejected 6 times over 3 years. Their feedback loop -- waiting ~6 months for feedback each time -- was painfully slow. We wanted to see if an agentic workflow can help researchers iterate faster.
When we trained the system on ICLR 2025 reviews and measured Spearman correlation (higher is better) on the test set:
- Correlation between two human reviewers: 0.41
- Correlation between AI and a human reviewer: 0.42
This suggests agentic reviewing is approaching human-level performance.
The agent grounds its feedback by searching arXiv, so it works best in fields like AI where research is freely published there. It’s an experimental tool, but I hope it helps you with your research.
Check it out here: paperreview.ai -
Andrew Ng
@AndrewYNg
After listening to multiple of Harry's 20VC podcasts, it was fun to appear on it. We talked about bottlenecks in AI adoption, US/China geopolitics, what opportunities remain to build in (tldr: a lot), and more!
Harry Stebbings
@HarryStebbings
“In my career working in AI, I have yet to meet a single AI person that ever felt like they had enough compute.”
“Open-weight models is a tremendous source of geopolitical influence”
“When China’s government makes an all-nation commitment, it’s a very powerful force I wouldn’t underestimate.”
Shows like this are why I love what I do.
My top 7 takeaways from one of the OGs of AI @AndrewYNg 👇
Spotify 👉 open.spotify.com/episode/7hNgy2ocZNpmMbqz8FHRla
Youtube 👉 www.youtube.com/watch?v=rT74mF6_NhQ
Apple Podcasts 👉 podcasts.apple.com/us/podcast/20vc-andrew-ng-on-the-bigge…
Timestamps:
00:00 Intro
01:04 What are the Biggest Bottlenecks in AI Today?
09:31 How LLMs Can Be Used as a Geopolitical Weapon
15:07 Should AI Talent Really Be Paid Billions?
19:15 Why is the Application Layer the Most Exciting Layer?
29:30 Will AI Deliver Masa Son's Predictions of 5% GDP Growth?
38:43 Do Margins Matter in a World of AI?
40:36 Is Defensibility Dead in a World of AI?
48:24 Will Human Labour Budgets Shift to AI Spend?
55:05 Are We in an AI Bubble?
56:28 Quick-Fire RoundThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I just got back from AI Dev x NYC, the AI developer conference where our community gathers for a day of coding, learning, and connecting. The vibe in the room was buzzing! It was at the last AI Dev in San Francisco that I met up with Kirsty Tan and started collaborating with her on what became our AI advisory firm AI Aspire. In-person meetings can spark new opportunities, and I hope the months to come will bring more stories about things that started in AI Dev x NYC!
The event was full of conversations about coding with AI, agentic AI, context engineering, governance, and building and scaling AI applications in startups and in large corporations. But the overriding impression I took away was one of near-universal optimism about our field, despite the mix of pessimism and optimism about AI in the broader world.
For example, many businesses have not yet gotten AI agents to deliver a significant ROI, and some AI skeptics are quoting an MIT study that said 95% of AI pilots are failing. (This study, by the way, has methodological flaws that make the viral headline misleading; see link in original post.) But at AI Dev were many of the teams responsible for the successful and rapidly growing set of AI applications. Speaking with fellow developers, I realized that because of AI's low penetration in businesses, it is simultaneously true that (a) many businesses do not yet have AI delivering significant ROI, and (b) many skilled AI teams are starting to deliver significant ROI and see the number of successful AI projects climbing rapidly, albeit from a low base. This is why AI developers are bullish about the growth that is to come.
Multiple exhibitors told me this was the best conference they had attended in a long time, because they got to speak with real developers. One told me that many other conferences seemed like fluff, whereas participants at AI Dev had much deeper technical understanding and thus were interested in and able to understand the nuances of cutting-edge technology. Whether the discussion was on observability of agentic workflows, the nuances of context engineering for AI coding, or a debate on how long the proliferation of RL gyms for training LLMs will continue, there was deep technical expertise in the room that lets us collectively see further into the future.
One special moment for me was when Nick Thompson, moderating a panel with Miriam Vogel and me, asked about governance. I replied that the United States’ recent hostile rhetoric toward immigrants is one of the worst moves it is making, and many in the audience clapped. Nick spoke about this moment in a video (links in original post).
I enjoyed meeting many people at AI Dev, and am grateful to everyone who came and to all our speakers, exhibitors, sponsors, volunteers, and event staff. My only regret is that, even though we scaled up the event 3x compared to the previous San Francisco event, we had to limit the number of tickets because the space couldn’t admit more attendees.
Even though so much work is now online, in-person events are special and can be turning points for individuals and projects. We plan to make the next AI Dev in San Francisco on April 28-29, 2026 an even bigger event, and look forward to the conference helping to spark more sharing and connections in the future.
[Original text: www.deeplearning.ai/the-batch/issue-328/ ] -
Andrew Ng
@AndrewYNg
I just dumped the latest NVIDIA 10-Q earnings report, released an hour ago, into Agentic Document Extraction, and the results are really accurate! Left side of the image shows the original PDF; right side shows the extracted info, including e.g. the $57.01B revenue in the most recent quarter. This is powered by the DPT (document pre-trained transformer) model. Check it out!
-
Andrew Ng
@AndrewYNg
New course announcement: Semantic Caching for AI Agents, taught by @tchutch94 and @ilzhechev from @Redisinc.
Semantic caching can significantly reduce your AI application's inference costs and latency. If someone asks "How do I get a refund?" and another later asks "I want my money back," semantic caching recognizes these mean the same thing so it can use a cached response instead of making another model call.
This short course takes you from building your first semantic cache from scratch to implementing production-ready systems using Redis' open-source tools.
Skills you'll gain:
- Build semantic caches from scratch, then implement them using Redis' SDK with production features
- Measure cache performance using hit rate, precision, recall, and latency
- Enhance accuracy with threshold tuning, cross-encoders, LLM validation, and fuzzy matching
Join and learn to reduce your agentic AI's costs and improve speed! www.deeplearning.ai/short-courses/semantic-caching-for-ai…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Really proud of the DeepLearningAI team. When Cloudflare went down, our engineers used AI coding to quickly implement a clone of basic Cloudflare capabilities to run our site on. So we came back up long before even major websites!
-
Andrew Ng
@AndrewYNg
I recently received an email titled “An 18-year-old’s dilemma: Too late to contribute to AI?” Its author, who gave me permission to share this, is preparing for college. He is worried that by the time he graduates, AI will be so good there’s no meaningful work left for him to do to contribute to humanity, and he will just live on Universal Basic Income (UBI). I wrote back to reassure him that there will still be plenty of work he can do for decades hence, and encouraged him to work hard and learn to build with AI. But this conversation struck me as an example of how harmful hype about AI is.
Yes, AI is amazingly intelligent, and I’m thrilled to be using it every day to build things I couldn’t have built a year ago. At the same time, AI is still incredibly dumb, and I would not trust a frontier LLM by itself to prioritize my calendar, carry out resumé screening, or choose what to order for lunch — tasks that businesses routinely ask junior personnel to do.
Yes, we can build AI software to do these tasks. For example, after a lot of customization work, one of my teams now has a decent AI resumé screening assistant. But the point is it took a lot of customization.
Even though LLMs can handle a much more general set of tasks than previous iterations of AI technology, compared to what humans can do, they are still highly specialized. They’re much better at working with text than other modalities, still require lots of custom engineering to get it the right context for a particular application, and we have few tools — and only inefficient ones — for getting our systems to learn from feedback and repeated exposure to a specific task (such as screening resumés for a particular role).
AI has stark limitations, and despite rapid improvements, it will remain limited compared to humans for a long time.
AI is amazing, but it has unfortunately been hyped up to be even more amazing than it is. A pernicious aspect of hype is that it often contains an element of truth, but not to the degree of the hype. This makes it difficult for nontechnical people to discern where the truth really is. Modern AI is a general purpose technology that is enabling many applications, but AI that can do any intellectual tasks that a human can (a popular definition for AGI) is still decades away or longer. This nuanced message that AI is general, but not that general, often is lost in the noise of today's media environment.
Similarly, the progress of frontier models is amazing! But not so amazing that they’ll be able to do everything under the sun without a lot of customization. I know VC investors who are scared to invest in application-layer startups because they are worried that frontier AI model companies will quickly wipe out all of these businesses by improving their models. While some thin wrappers around LLMs no doubt will be replaced, there also remains a huge set of valuable applications that the current trajectory of progress of frontier models won’t displace for a long time.
Without accurate information about the current state of AI and how it is likely to progress, some young people will decide not to enter AI because think think AGI leaves them no meaningful role, or decide not to learn how to code because they fear AI will automate it — right when it is the best time ever to join our field.
Let us all keep working to get to a precise understanding of what’s actually possible, and keep building!
[Original text: www.deeplearning.ai/the-batch/issue-327/ ] -
Andrew Ng
@AndrewYNg
New course announcement: Design, Develop, and Deploy Multi-Agent Systems with CrewAI, taught by @joaomdmoura, @crewAIInc Co-founder and CEO.
Multi-agent systems let you build AI teams that work together to automate complex workflows, similar to how human teams work.
CrewAI makes it simple to build multi-agent systems that handle routine work for you—just define your agents, tasks, and crew, and it manages the complexity of coordinating multiple agents and their context automatically. (Disclosure: I made a small angel investment in CrewAI.) This course takes you from building your first agent to deploying production systems using the open-source CrewAI framework.
Skills you'll gain:
- Build reliable AI agents equipped with tools, memory, and guardrails
- Develop teams of agents that can plan, reason, and coordinate
- Deploy production-ready systems with tracing, evaluation, and monitoring
Whether you’re exploring multi-agent systems for the first time or looking to take your projects further, this course will help you build a mental framework for designing multi-agent systems, and help you turn ideas into scalable, production-ready applications.
Sign up here: www.deeplearning.ai/courses/design-develop-and-deploy-mul…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI agents are getting better at looking at different types of data in businesses to spot patterns and create value. This is making data silos increasingly painful. This is why I increasingly try to select software that lets me control my own data, so I can make it available to my AI agents.
Because of AI’s growing capabilities, the value you can now create from “connecting the dots” between different pieces of data is higher than ever. For example, if an email click is logged in one vendor’s system and a subsequent online purchase is logged in a different one, then it is valuable to build agents that can access both of these data sources to see how they correlate to make better decisions.
Unfortunately, many SaaS vendors try to create a data silo in their customer’s business. By making it hard for you to extract your data, they create high switching costs. This also allows them to steer you to buy their AI agent services — sometimes at high expense and/or of low quality — rather than build your own or buy from a different vendor. Unfortunately, some SaaS vendors are seeing AI agents coming for this data and working to make it harder for you (and your AI agents) to efficiently access it.
One of my teams just told me that a SaaS vendor we have been using to store our customer data wants to charge over $20,000 for an API key to get at our data. This high cost — no doubt intentionally designed to make it hard for customers to get their data out — is adding a barrier to implementing agentic workflows that take advantage of that data.
Through AI Aspire (an AI advisory firm), I advise a number of businesses on their AI strategies. When it comes to buying SaaS, I often advise them to try to control their own data (which, sadly, some vendors mightily resist). This way, you can hire a SaaS vendor to record and operate on your data, but ultimately you decide how to route it to the appropriate human or AI system for processing.
Over the past decade, a lot of work has gone into organizing businesses’ structured data. Because AI can now process unstructured data much better than before, the value of organizing your unstructured data (including PDF files, which LandingAI’s Agentic Document Extraction specializes in!) is higher than ever before.
In the era of generative AI, businesses and individuals have important work ahead to organize their data to be AI-ready.
P.S. As an individual, my favorite note-taking app is Obsidian. I am happy to “hire” Obsidian to operate on my notes files. And, all my notes are saved as Markdown files in my file system, and I have built AI agents that read from or write to my Obsidian files. This is a small example of how controlling my own notes data lets me do more with AI agents!
[Original text: www.deeplearning.ai/the-batch/issue-326/ ] -
Andrew Ng
@AndrewYNg
.@roelofbotha is one of the investors I've come to most respect, and I learned something new every time I spoke with him. Even beyond his leadership of @sequoia, his influence on how investors think is hard to overstate. His passing on the baton is the end of an era!
-
Andrew Ng
@AndrewYNg
AI coding just arrived in Jupyter notebooks - and @brganger (Jupyter co-founder) and I will show you how to use it.
Coding by hand is becoming obsolete. The latest Jupyter AI - built by the Jupyter team and showcased at JupyterCon this week - brings AI assistance directly into notebooks.
Most AI coding assistants struggle with Jupyter notebooks. Jupyter AI was designed specifically for them. This is the first course to teach it.
In this course, Brian and I teach you to:
- Generate and debug code directly in notebook cells through an integrated chat interface
- Provide the right context (like API docs) to help AI write accurate code
- Use Jupyter AI's unique notebook features: drag cells to chat, generate cells from chat, attach context for the LLM
We've integrated Jupyter AI directly into the DeepLearningAI platform, so you can start using it immediately. Since Jupyter AI is open source, you can also install and run it locally afterward.
Whether you're experienced with notebooks or learning them for the first time, this course will prepare you for AI-assisted notebook development.
Start using Jupyter AI (free): www.deeplearning.ai/short-courses/jupyter-ai-coding-in-no…This media couldn't be displayed. See it on the original post.
Andrew Ng
@AndrewYNg
Wrong tag - I’m so sorry. This is co-taught with the wonderful Brian Granger @ellisonbg !
-
Andrew Ng
@AndrewYNg
Haha! Good one.
Christopher Manning
@chrmanning
We already called it in 2020!
-
Andrew Ng
@AndrewYNg
deeplearning.ai Pro is now generally available -- this is the one membership that keeps you at the forefront of AI. Please join!
There has never been a moment when the distance between having an idea and building it has been smaller. Things that required months of work for teams can now be built by individuals using AI, in days. This is why we built deeplearning.ai Pro. I'm personally working hard on this membership program to help you to build applications that can launch or accelerate your career, and shape the future of AI.
deeplearning.ai Pro gives you full access to 150+ programs, including my recently launched Agentic AI course, the new Post-Training and PyTorch courses by Sharon Zhou and Laurence Moroney (just released this week), and all of deeplearning.ai's top courses and professional certificates.
All course videos remain free. Pro membership adds hands-on learning: labs to build working systems, practice questions to hone your understanding, and certificates to share your skills.
I'm also building new tools to help you create AI applications and grow your career (and have fun doing so!). Many will be available first to Pro members.
Try out deeplearning.ai Pro free, and let me know what you build!
Join here: learn.deeplearning.ai/membership -
Andrew Ng
@AndrewYNg
In hindsight the importance of GPUs for AI was something we really got right!
illiquid
@lefttailguy
NVIDIA $5T today, but the writing has been on the wall Since @AndrewYNg and team published this paper in 2009.
Were there any large/notable hedge funds that actually nailed the NVIDIA trade early? Say 2016/2017? There was ALOT of writing on the wall at that point, and company was big enough such that larger funds could place a scaled bet. Why did they miss it?
Massive blunder if you consider yourself a fundamentally oriented "technology investor".... -
Andrew Ng
@AndrewYNg
An exciting new professional certificate: PyTorch for Deep Learning taught by @lmoroney is now available at deeplearning.ai. This is the definitive program for learning PyTorch, which is one of the main frameworks researchers use to build breakthrough AI systems. If you want to understand how modern deep learning models work—or build your own custom architectures—PyTorch gives you direct control over the key aspects of model development.
This three-course professional certificate takes you from fundamentals through advanced architectures and deployment:
Course 1: PyTorch: Fundamentals - Learn how PyTorch represents data with tensors and how datasets fit into the training process. You'll build and train neural networks step by step, monitor training progress, and evaluate performance. By the end, you'll understand PyTorch's workflow and be ready to design, train, and test your own models.
Course 2: PyTorch: Techniques and Ecosystem Tools - Master hyperparameter optimization, model profiling, and workflow efficiency. You'll use learning rate schedulers, tackle overfitting, and apply automated tuning with Optuna. Work with TorchVision for visual AI and Hugging Face for NLP. Learn transfer learning and fine-tune pretrained models for new problems.
Course 3: PyTorch: Advanced Architectures and Deployment - Build sophisticated architectures including Siamese Networks, ResNet, DenseNet, and Transformers. Learn how attention mechanisms power modern language models and how diffusion models generate images. Prepare models for deployment with ONNX, MLflow, pruning, and quantization.
Skills you'll gain:
- Build and optimize neural networks in PyTorch—the framework researchers use to create breakthrough models
- Fine-tune pretrained models for computer vision and NLP tasks—adapting existing models to solve your specific problems
- Implement transformer architectures and work with diffusion models, the core technologies behind ChatGPT and modern image generation
- Optimize models with quantization and pruning to make them fast and efficient for real-world deployment
Whether you want to use pre-existing models, build your own custom models, or just understand what's happening under the hood of the systems you use, this specialization will give you that foundation.
Start learning PyTorch: www.deeplearning.ai/courses/pytorch-for-deep-learning-pro…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
An exciting new course: Fine-tuning and Reinforcement Learning for LLMs: Intro to Post-training, taught by @realSharonZhou, VP of AI at @AMD. Available now at deeplearning.ai.
Post-training is the key technique used by frontier labs to turn a base LLM--a model trained on massive unlabeled text to predict the next word/token--into a helpful, reliable assistant that can follow instructions. I've also seen many applications where post-training is what turns a demo application that works only 80% of the time into a reliable system that consistently performs. This course will teach you the most important post-training techniques!
In this 5 module course, Sharon walks you through the complete post-training pipeline: supervised fine-tuning, reward modeling, RLHF, and techniques like PPO and GRPO. You'll also learn to use LoRA for efficient training, and to design evals that catch problems before and after deployment.
Skills you'll gain:
- Apply supervised fine-tuning and reinforcement learning (RLHF, PPO, GRPO) to align models to desired behaviors
- Use LoRA for efficient fine-tuning without retraining entire models
- Prepare datasets and generate synthetic data for post-training
- Understand how to operate LLM production pipelines, with go/no-go decision points and feedback loops
These advanced methods aren’t limited to frontier AI labs anymore, and you can now use them in your own applications.
Learn here: deeplearning.ai/courses/fine-tuning-and-reinforcement-lea…This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
LandingAI
@LandingAI
The AI Financial Hackathon Championship is officially live! 🚀
We’re hosting this in collaboration with @awscloud (AWS), bringing together builders who are pushing the limits of intelligent document processing.
This is your chance to build and design solutions that turn complex financial documents into structured, actionable data using Agentic Document Extraction (ADE).
To participate:
• Register before November 9 at 11:59 PM ET and start building right away
• Submit your project by November 10 for a chance to qualify for the live finale in New York on November 15
• The top 10 teams will be invited to present their projects in person
💰 Prize Pool
1st Place: $5,000 ($3K cash + $2K AWS credits)
2nd Place: $1,000
Best Online-Only App: Surprise prize shipped to the team
👉 Register here: luma.com/jme15h1t -
Andrew Ng
@AndrewYNg
Hanging out with Project Jupyter co-founder @ellisonbg. If not for him and @fperez_org we wouldn’t have the coding notebooks we use daily in AI and Data Science. Very grateful to him and the whole Jupyter team for this wonderful open-source work!
-
Andrew Ng
@AndrewYNg
The full agenda for AI Dev 25 x NYC is ready.
Developers from Google, AWS, Vercel, Groq, Mistral AI, SAP, and other exciting companies will share what they've learned building production AI systems. Here's what we'll cover:
Agentic Architecture: When orchestration frameworks help versus when they accumulate errors. How model-driven agents and autonomous planning handle edge cases.
Context Engineering: Why retrieval fails for complex reasoning tasks. How knowledge graphs connect information that vector search misses. Building memory systems that preserve relationships.
Infrastructure: Where hardware, models, and applications create scaling bottlenecks. Semantic caching strategies that cut costs and latency. How inference speed enables better orchestration.
Production Readiness: Moving from informal evaluation to systematic agent testing. Translating AI governance into engineering practice. Building under regulatory constraints.
Tooling: MCP implementations that work. Context-rich code review systems. Working demos you can adapt for your applications.
I'll share my perspective on where AI development is heading. Looking forward to seeing you there! ai-dev.deeplearning.ai -
Andrew Ng
@AndrewYNg
Without proper governance, an AI agent might autonomously access sensitive data, expose personal information, or modify sensitive records. In our new short course: “Governing AI Agents,” created with @Databricks and taught by Amber Roberts, you’ll design AI agents that handle data safely, securely, and transparently across their entire lifecycle.
You’ll learn to integrate governance into your agent’s workflow by controlling data access, ensuring privacy protection and implementing observability.
Skills you'll gain:
- Understand the four pillars of agent governance: Lifecycle management, risk management, security, and observability
- Define appropriate data permissions for your agent
- Create views or SQL queries that return only the data your agent should access
- Anonymize and mask sensitive data like social security numbers and employee IDs
- Log, evaluate, version, and deploy your agents on Databricks
If you’re building or deploying AI agents, learning how to govern them is key to keeping systems safe and production-ready.
Sign up here: www.deeplearning.ai/short-courses/governing-ai-agentsThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Fun breakfast with @ylecun. We chatted about open science and open source (grateful for his tireless advocacy of these for decades), JEPA and where AI research and models might go next!
-
Andrew Ng
@AndrewYNg
Readers responded with both surprise and agreement last week when I wrote that the single biggest predictor of how rapidly a team makes progress building an AI agent lay in their ability to drive a disciplined process for evals (measuring the system’s performance) and error analysis (identifying the causes of errors). It’s tempting to shortcut these processes and to quickly attempt fixes to mistakes rather than slowing down to identify the root causes. But evals and error analysis can lead to much faster progress. In this first of a two-part letter, I’ll share some best practices for finding and addressing issues in agentic systems.
Even though error analysis has long been an important part of building supervised learning systems, it is still underappreciated compared to, say, using the latest and buzziest tools. Identifying the root causes of particular kinds of errors might seem “boring,” but it pays off! If you are not yet persuaded that error analysis is important, permit me to point out:
- To master a composition on a musical instrument, you don’t only play the same piece from start to end. Instead, you identify where you’re stumbling and practice those parts more.
- To be healthy, you don’t just build your diet around the latest nutrition fads. You also ask your doctor about your bloodwork to see if anything is amiss. (I did this last month and am happy to report I’m in good health! 😃)
- To improve your sports team’s performance, you don’t just practice trick shots. Instead, you review game films to spot gaps and then address them.
To improve your agentic AI system, don’t just stack up the latest buzzy techniques that just went viral on social media (though I find it fun to experiment with buzzy AI techniques as much as the next person!). Instead, use error analysis to figure out where it’s falling short, and focus on that.
Before analyzing errors, we first have to decide what is an error. So the first step is to put in evals. I’ll focus on that for the remainder of this letter and discuss error analysis next week.
If you are using supervised learning to train a binary classifier, the number of ways the algorithm could make a mistake is limited. It could output 0 instead of 1, or vice versa. There is also a handful of standard metrics like accuracy, precision, recall, F1, ROC, etc. that apply to many problems. So as long as you know the test distribution, evals are relatively straightforward, and much of the work of error analysis lies in identifying what types of input an algorithm fails on, which also leads to data-centric AI techniques for acquiring more data to augment the algorithm in areas where it’s weak.
With generative AI, a lot of intuitions from evals and error analysis of supervised learning carry over — history doesn’t repeat itself, but it rhymes — and developers who are already familiar with machine learning and deep learning often adapt to generative AI faster than people who are starting from scratch. But one new challenge is that the space of outputs is much richer, so there are many more ways an algorithm’s output might be wrong.
Take the example of automated processing of financial invoices where we use an agentic workflow to populate a financial database with information from received invoices. Will the algorithm incorrectly extract the invoice due date? Or the final amount? Or mistake the payer address for the biller address? Or get the financial currency wrong? Or make the wrong API call so the verification process fails? Because the output space is much larger, the number of failure modes is also much larger.
Rather than defining an error metric ahead of time, it is therefore typically more effective to first quickly build a prototype, then manually examine a handful of agent outputs to see where it performs well and where it stumbles. This allows you to focus on building datasets and error metrics — sometimes objective metrics implemented in code, and sometimes subjective metrics using LLM-as-judge — to check the system’s performance in the dimensions you are most concerned about. In supervised learning, we sometimes tune the error metric to better reflect what humans care about. With agentic workflows, I find tuning evals to be even more iterative, with more frequent tweaks to the evals to capture the wider range of things that can go wrong.
I discuss this and other best practices in detail in Module 4 of the Agentic AI course on deeplearning.ai that we announced last week. After building evals, you now have a measurement of your system’s performance, which provides a foundation for trying different modifications to your agent, as you can now measure what makes a difference. The next step is then to perform error analysis to pinpoint what changes to focus your development efforts on. I’ll discuss this further next week.
[Original text: www.deeplearning.ai/the-batch/issue-323/ ] -
Andrew Ng
@AndrewYNg
Learn to build your own voice-activated AI assistant that can execute tasks like gathering recent AI news from the web, scripting out a podcast, and using tools to put all that into a multi-speaker podcast. See our new short course: "Building Live Voice Agents with Google’s ADK (Agent Development Kit),” taught by Google’s @lavinigam and @sitalakshmi_s.
ADK provides modular components that make it easy to build and debug agents. It also includes a built-in web interface for tracing agentic reasoning. This course illustrates these concepts via building a live voice agent that can chain actions to complete a complex task like creating a podcast. This requires maintaining context, implementing guardrails, reasoning, and handling audio streaming, while keeping latency low.
You’ll learn to:
- Build voice agents that listen, reason, and respond
- Guide your agent to follow a specific workflow to accomplish a task
- Coordinate specialized agents to build an agentic podcast workflow that researches topics and produces multi-speaker audio
- Understand how to deploy an agent into production
Even if you’re not yet building voice systems, you'll find understanding how realtime agents stream data and maintain reliability useful for designing modern agentic applications.
Please join here: www.deeplearning.ai/short-courses/building-live-voice-age…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Announcing my new course: Agentic AI!
Building AI agents is one of the most in-demand skills in the job market. This course, available now at deeplearning.ai, teaches you how.
You'll learn to implement four key agentic design patterns:
- Reflection, in which an agent examines its own output and figures out how to improve it
- Tool use, in which an LLM-driven application decides which functions to call to carry out web search, access calendars, send email, write code, etc.
- Planning, where you'll use an LLM to decide how to break down a task into sub-tasks for execution, and
- Multi-agent collaboration, in which you build multiple specialized agents — much like how a company might hire multiple employees — to perform a complex task
You'll also learn to take a complex application and systematically decompose it into a sequence of tasks to implement using these design patterns.
But here's what I think is the most important part of this course: Having worked with many teams on AI agents, I've found that the single biggest predictor of whether someone executes well is their ability to drive a disciplined process for evals and error analysis. In this course, you'll learn how to do this, so you can efficiently home in on which components to improve in a complex agentic workflow. Instead of guessing what to work on, you'll let evals data guide you. This will put you significantly ahead of the game compared to the vast majority of teams building agents.
Together, we'll build a deep research agent that searches, synthesizes, and reports, using all of these agentic design patterns and best practices.
This self-paced course is taught in a vendor neutral way, using raw Python - without hiding details in a framework. You'll see how each step works, and learn the core concepts that you can then implement using any popular agentic AI framework, or using no framework. The only prerequisite is familiarity with Python, though knowing a bit about LLMs helps.
Come join me, and let's build some agentic AI systems!
Sign up to get started: www.deeplearning.ai/courses/agentic-ai/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Announcing a significant upgrade to Agentic Document Extraction!
LandingAI's new DPT (Document Pre-trained Transformer) accurately extracts even from complex docs. For example, from large, complex tables, which is important for many finance and healthcare applications. And a new SDK makes using it require only 3 simple lines of code. Please see the video for technical details. I hope this unlocks a lot of value from the "dark data" currently stuck in PDF files, and that you'll build something cool with this!This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Last week, China barred its major tech companies from buying Nvidia chips. This move received only modest attention in the media, but has implications beyond what’s widely appreciated. Specifically, it signals that China has progressed sufficiently in semiconductors to break away from dependence on advanced chips designed in the U.S., the vast majority of which are manufactured in Taiwan. It also highlights the U.S. vulnerability to possible disruptions in Taiwan at a moment when China is becoming less vulnerable.
After the U.S. started restricting AI chip sales to China, China dramatically ramped up its semiconductor research and investment to move toward self-sufficiency. These efforts are starting to bear fruit, and China’s willingness to cut off Nvidia is a strong sign of its faith in its domestic capabilities. For example, the new DeepSeek-R1-Safe model was trained on 1000 Huawei Ascend chips. While individual Ascend chips are significantly less powerful than individual Nvidia or AMD chips, Huawei’s system-level design approach to orchestrating how a much larger number of chips work together seems to be paying off. For example, Huawei’s CloudMatrix 384 system of 384 chips aims to compete with Nvidia’s GB200, which uses 72 higher-capability chips.
Today, U.S. access to advanced semiconductors is heavily dependent on Taiwan’s TSMC, which manufactures the vast majority of the most advanced chips. Unfortunately, U.S. efforts to ramp up domestic semiconductor manufacturing have been slow. I am encouraged that one fab at the TSMC Arizona facility is now operating, but issues of workforce training, culture, licensing and permitting, and the supply chain are still being addressed, and there is still a long road ahead for the U.S. facility to be a viable substitute for manufacturing in Taiwan.
If China gains independence from Taiwan manufacturing significantly faster than the U.S., this would leave the U.S. much more vulnerable to possible disruptions in Taiwan, whether through natural disasters or man-made events. If manufacturing in Taiwan is disrupted for any reason and Chinese companies end up accounting for a large fraction of global semiconductor manufacturing capabilities, that would also help China gain tremendous geopolitical influence.
Despite occasional moments of heightened tensions and large-scale military exercises, Taiwan has been mostly peaceful since the 1960s. This peace has helped the people of Taiwan to prosper and allowed AI to make tremendous advances, built on top of chips made by TSMC. I hope we will find a path to maintaining peace for many decades more.
But hope is not a plan. In addition to working to ensure peace, practical work lies ahead to multi-source, build more chip fabs in more nations, and enhance the resilience of the semiconductor supply chain. Dependence on any single manufacturer invites shortages, price spikes, and stalled innovation the moment something goes sideways.
[Original text: www.deeplearning.ai/the-batch/issue-320/ ] -
Andrew Ng
@AndrewYNg
When data agents fail, they often fail silently - giving confident-sounding answers that are wrong, and it can be hard to figure out what caused the failure.
"Building and Evaluating Data Agents" is a new short course created with @Snowflake and taught by @datta_cs and @_jreini that teaches you to build data agents with comprehensive evaluation built in.
Skills you'll gain:
- Build reliable LLM data agents using the Goal-Plan-Action framework and runtime evaluations that catch failures mid-execution
- Use OpenTelemetry tracing and evaluation infrastructure to diagnose exactly where agents fail and systematically improve performance
- Orchestrate multi-step workflows across web search, SQL, and document retrieval in LangGraph-based agents
The result: visibility into every step of your agent's reasoning, so if something breaks, you have a systematic approach to fix it.
Sign up to get started: www.deeplearning.ai/short-courses/building-and-evaluating…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
My heart goes out to all the families and individuals anxious over their futures following the abrupt and chaotic announcement of H-1B visa changes.
America should be working to attract more skilled talent, not create uncertainly that turns them away. To all legal immigrants and H1-B holders: I support and appreciate you. -
Andrew Ng
@AndrewYNg
Automated software testing is growing in importance in the era of AI-assisted coding. Agentic coding systems accelerate development but are also unreliable. Agentic testing — where you ask AI to write tests and check your code against them — is helping. Automatically testing infrastructure software components that you intend to build on top of is especially helpful and results in more stable infrastructure and less downstream debugging.
Software testing methodologies such as Test Driven Development (TDD), a test-intensive approach that involves first writing rigorous tests for correctness and only then making progress by writing code that passes those tests, are an important way to find bugs. But it can be a lot of work to write tests. (I personally never adopted TDD for that reason.) Because AI is quite good at writing tests, agentic testing enjoys growing attention.
First, coding agents do misbehave! My teams use them a lot, and we have seen:
- Numerous bugs introduced by coding agents, including subtle infrastructure bugs that take humans weeks to find.
- A security loophole that was introduced into our production system when a coding agent made password resets easier to simplify development.
- Reward hacking, where a coding agent modified test code to make it easier to pass the tests.
- An agent running "rm *.py" in the working directory, leading to deletion of all of a project's code (which, fortunately, was backed up on github).
In the last example, when pressed, the agent apologized and agreed “that was an incredibly stupid mistake.” This made us feel better, but the damage had already been done!
I love coding agents despite such mistakes and see them making us dramatically more productive. To make them more reliable, I’ve found that prioritizing where to test helps.
I rarely write (or direct an agent to write) extensive tests for front-end code. If there's a bug, hopefully it will be easy to see and also cause little lasting damage. For example, I find generated code’s front-end bugs, say in the display of information on a web page, relatively easy to find. When the front end of a web site looks wrong, you’ll see it immediately, and you can tell the agent and have it iterate to fix it. (A more advanced technique: Use MCP to let the agent integrate with software like Playwright to automatically take screenshots, so it can autonomously see if something is wrong and debug.)
In contrast, back-end bugs are harder to find. I’ve seen subtle infrastructure bugs — for example, one that led to a corrupted database record only in certain corner cases — that took a long time to find. Putting in place rigorous tests for your infrastructure code might help spot these problems earlier and save you many hours of challenging debugging.
Bugs in software components that you intend to build on top of lead to downstream bugs that can be hard to find. Further, bugs in a component that’s deep in a software stack — and that you build multiple abstraction layers on top of — might surface only weeks or months later, long after you’ve forgotten what you were doing while building this specific component, and be really hard to identify and fix. This is why testing components deep in your software stack is especially important. Meta’s mantra “Move fast with stable infrastructure” (which replaced “move fast and break things”) still applies today. Agentic testing can help you make sure you have good infrastructure for you and others to build on!
At AI Fund and deeplearning.ai’s recent Buildathon, we held a panel discussion with experts in agentic coding (Michele Catasta, President at Replit; Chao Peng, Principal Research Scientist at Trae; and Paxton Maeder-York, Venture Partnerships at Anthropic; moderated by AI Fund’s Eli Chen), where the speakers shared best practices. Testing was one of the topics discussed. That panel was one of my highlights of Buildathon and you can watch the video on YouTube.
[Original text: www.deeplearning.ai/the-batch/issue-319/ ]
Andrew Ng
@AndrewYNg
Video of the panel at Buildathon: www.youtube.com/watch?v=9VxB8ewCHN0
-
Andrew Ng
@AndrewYNg
New short course: Build AI Apps with MCP Servers: Working with Box Files, built with @Box and taught by @BenAtBox , their CTO.
Many AI applications require custom code for basic file operations. The Model Context Protocol (MCP) standardizes this by letting you offload file tasks to dedicated servers that provide tools an LLM can use directly.
In this course, you'll process documents stored in a Box folder using the Box MCP server. Rather than writing custom integration code to connect to the Box API and download files, you'll design your application to use the tools provided via MCP.
Skills you'll gain:
- Build an LLM-powered document processing app, using the Box MCP server to access files
- Design a multi-agent system using Google's Agent Development Kit (ADK), consisting of specialized agents for file operations
- Coordinate the multi-agent workflow through an orchestrator that uses the Agent2Agent (A2A) protocol to connect to the agents
You'll start with a local file-processing app, refactor it to work with Box's MCP server, then evolve it into a multi-agent system.
Sign up here: www.deeplearning.ai/short-courses/build-ai-apps-with-mcp-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
There is significant unmet demand for developers who understand AI. At the same time, because most universities have not yet adapted their curricula to the new reality of programming jobs being much more productive with AI tools, there is also an uptick in unemployment of recent CS graduates.
When I interview AI engineers — people skilled at building AI applications — I look for people who can:
- Use AI assistance to rapidly engineer software systems
- Use AI building blocks like prompting, RAG, evals, agentic workflows, and machine learning to build applications
- Prototype and iterate rapidly
Someone with these skills can get a massively greater amount done than someone who writes code the way we did in 2022, before the advent of Generative AI. I talk to large businesses every week that would love to hire hundreds or more people with these skills, as well as startups that have great ideas but not enough engineers to build them. As more businesses adopt AI, I expect this talent shortage only to grow! At the same time, recent CS graduates face an increased unemployment rate, though the underemployment rate — of graduates doing work that doesn’t require a degree — is still lower than for most other majors. This is why we hear simultaneously anecdotes of unemployed CS graduates and also of rising salaries for in-demand AI engineers.
When programming evolved from punchcards to keyboard and terminal, employers continued to hire punchcard programmers for a while. But eventually, all developers had to switch to the new way of coding. AI engineering is similarly creating a huge wave of change.
There is a stereotype of “AI Native” fresh college graduates who outperform experienced developers. There is some truth to this. Multiple times, I have hired, for full-stack software engineering, a new grad who really knows AI over an experienced developer who still works 2022-style. But the best developers I know aren’t recent graduates (no offense to the fresh grads!). They are experienced developers who have been on top of changes in AI. The most productive programmers today deeply understand computers, how to architect software, and how to make complex tradeoffs — and who additionally are familiar with cutting-edge AI tools.
Sure, some skills from 2022 are becoming obsolete. For example, a lot of coding syntax that we had to memorize back then is no longer important, since we no longer need to code by hand as much. But even if, say, 30% of CS knowledge is obsolete, the remaining 70% — complemented with modern AI knowledge — is what makes really productive developers. (Even after punch cards became obsolete, a fundamental understanding of programming was very helpful for typing code into a keyboard.)
Without understanding how computers work, you can’t just “vibe code” your way to greatness. Fundamentals are still important, and for those who additionally understand AI, job opportunities are numerous!
[Original text: www.deeplearning.ai/the-batch/issue-317/ ] -
Andrew Ng
@AndrewYNg
Parallel agents are emerging as an important new direction for scaling up AI. AI capabilities have scaled with more training data, training-time compute, and test-time compute. Having multiple agents run in parallel is growing as a technique to further scale and improve performance.
We know from work at Baidu by my former team, and later OpenAI, that AI models’ performance scales predictably with the amount of data and training computation. Performance rises further with test-time compute such as in agentic workflows and in reasoning models that think, reflect, and iterate on an answer. But these methods take longer to produce output. Agents working in parallel offer another path to improve results, without making users wait.
Reasoning models generate tokens sequentially and can take a long time to run. Similarly, most agentic workflows are initially implemented in a sequential way. But as LLM prices per token continue to fall — thus making these techniques practical — and product teams want to deliver results to users faster, more and more agentic workflows are being parallelized.
Some examples:
- Many research agents now fetch multiple web pages and examine their texts in parallel to try to synthesize deeply thoughtful research reports more quickly.
- Some agentic coding frameworks allow users to orchestrate many agents working simultaneously on different parts of a code base. Our short course on Claude Code shows how to do this using git worktrees.
- A rapidly growing design pattern for agentic workflows is to have a compute-heavy agent work for minutes or longer to accomplish a task, while another agent monitors the first and gives brief updates to the user to keep them informed. From here, it’s a short hop to parallel agents that work in the background while the UI agent keeps users informed and perhaps also routes asynchronous user feedback to the other agents.
It is difficult for a human manager to take a complex task (like building a complex software application) and break it down into smaller tasks for human engineers to work on in parallel; scaling to huge numbers of engineers is especially challenging. Similarly, it is also challenging to decompose tasks for parallel agents to carry out. But the falling cost of LLM inference makes it worthwhile to use a lot more tokens, and using them in parallel allows this to be done without significantly increasing the user’s waiting time.
I am also encouraged by the growing body of research on parallel agents. For example, I enjoyed reading “CodeMonkeys: Scaling Test-Time Compute for Software Engineering” by Ryan Ehrlich and others, which shows how parallel code generation helps you to explore the solution space. The mixture-of-agents architecture by Junlin Wang is a surprisingly simple way to organize parallel agents: Have multiple LLMs come up with different answers, then have an aggregator LLM combine them into the final output.
There remains a lot of research as well as engineering to explore how best to leverage parallel agents, and I believe the number of agents that can work productively in parallel — like the humans who can work productively in parallel — will be very high.
[Original text, with links: www.deeplearning.ai/the-batch/issue-316/ ] -
Andrew Ng
@AndrewYNg
Build better RAG by letting a team of agents extract and connect your reference materials into a knowledge graph. Our new short course, “Agentic Knowledge Graph Construction,” taught by @Neo4j Innovation Lead @akollegger, shows you how.
Knowledge graphs are an important way to store information accurately but they are a lot of work to build manually.
In this course you’ll learn how to build a team of agents that turn data– in this case product reviews and invoices from suppliers–into structured graphs of entities and relationships for RAG.
Learn how agents can automatically handle the time-consuming work of building graphs — extracting entities and relationships (e.g., Product "contains" Assembly, Part "supplied_by" Supplier, Customer review "mentions" Product), deduplicating them, fact-checking them, and committing them to a graph database — so your retrieval system can find right information to generate accurate output. For example, you can use agents to help trace customer complaints directly to specific suppliers, manufacturing processes, and product hierarchies, thus turning fragmented information into queryable business intelligence.
Skills you’ll gain:
- Build, store, and access knowledge graphs using the Neo4j graph database
- Build multi-agent systems using Google’s Agent Development Kit (ADK)
- Set up a loop of agentic workflows to propose and refine a graph schema through fact-checking
- Connect agent-generated graphs of unstructured and structured data into a unified knowledge graph
This course gets into the practicum of why knowledge graphs give more accurate information retrieval than vector search alone, especially for high-stakes applications where precision matters more than fuzzy similarity matching.
Sign up here: www.deeplearning.ai/short-courses/agentic-knowledge-graph…This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Niklas Muennighoff
@Muennighoff
Can AI solve open problems in math, physics, coding, medical sciences & beyond?
We collected unsolved questions (UQ) & tested frontier LLMs. Some solutions passed expert validation… -
Andrew Ng reposted
Elad Gil
@eladgil
Enjoyed chatting w @AndrewYNg
* Bleeding Edge of Agentic AI
* Will Models Bootstrap Themselves?
* Vibe Coding vs. AI Assisted Coding
* Successful Founder Profiles
* Next Industry Transformations
www.youtube.com/watch?v=SYisFbhR7xs&list=PLMKa0PxGwad7jf8… -
Andrew Ng reposted
sarah guo
@saranormous
my longtime friend, AI legend @AndrewYNg joined me and @eladgil on @NoPriorsPod this week -- he thinks AI can really help with investing, but it can't get at some of the people judgement or transparency borne of relationships (yet)
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
On Saturday at the Buildathon hosted by AI Fund and deeplearning.ai, over 100 developers competed to build software products quickly using AI assisted coding. I was inspired to see developers build functional products in just 1-2 hours. The best practices for rapid engineering are changing quickly along with the tools, and I loved the hallway conversations sharing tips with other developers on using AI to code!
The competitors raced to fulfill product specs like this one (you can see the full list in our github repo; link in reply):
Project: Codebase Time Machine
Description: Navigate any codebase through time, understanding evolution of features and architectural decisions.
Requirements:
- Clone repo and analyze full git history
- Build semantic understanding of code changes over time
- Answer questions like “Why was this pattern introduced?” or “Show me how auth evolved”
- Visualize code ownership and complexity trends
- Link commits to business features/decisions
Teams had 6½ hours to build 5 products. And many of them managed to do exactly that! They created fully functional applications with good UIs and sometimes embellishments.
What excites me most isn’t just what can now be built in a few hours. Rather, it is that, if AI assistance lets us build basic but fully functional products this quickly, then imagine what can now be done in a week, or a month, or six months. If the teams that participated in the Buildathon had this velocity of execution and iterated over multiple cycles of getting customer feedback and using that to improve the product, imagine how quickly it is now possible to build great products.
Owning proprietary software has long been a moat for businesses, because it has been hard to write complex software. Now, as AI assistance enables rapid engineering, this moat is weakening.
While many members of the winning teams had computer science backgrounds — which does provide an edge — not all did. Team members who took home prizes included a high school senior, a product manager, and a healthcare entrepreneur who initially posted on Discord that he was “over his skis” as someone who “isn't a coder.” I was thrilled that multiple participants told me they exceeded their own expectations and discovered they can now build faster than they realized. If you haven’t yet pushed yourself to build quickly using agentic coding tools, you, too, might be surprised at what you can do!
At AI Fund and deeplearning.ai, we pride ourselves on building and iterating quickly. At the Buildathon, I saw many teams execute quickly using a wide range of tools including Claude Code, GPT-5, Replit, Cursor, Windsurf, Trae, and many others.
I offer my hearty congratulations to all the winners!
- 1st Place: Milind Pathak, Mukul Pathak, and Sapna Sangmitra (Team Vibe-as-a-Service), a team of three family members. They also received an award for Best Design.
- 2nd Place: David Schuster, Massimiliano Viola, and Manvik Pasula. (Team Two Coders and a Finance Guy).
- Solo Participant Award: Ivelina Dimova, who had just flown to San Francisco from Portugal, and who worked on the 5 projects not sequentially, but in parallel!
- Graph Thinking Award: Divya Mahajan, Terresa Pan, and Achin Gupta (Team A-sync).
- Honorable mentions went to finalists Alec Hewitt, Juan Martinez, Mark Watson and Sophia Tang (Team Secret Agents) and Yuanyuan Pan, Jack Lin, and Xi Huang (Team Can Kids).
To everyone who participated, thank you! Through events like these, I hope we can all learn from each other, encourage each other, invent new best practices, and spread the word about where agentic coding is taking software engineering.
[Original text: www.deeplearning.ai/the-batch/issue-315/ ]
Andrew Ng
@AndrewYNg
The products that teams worked on at Buildathon: github.com/AIFundTeam/buildathon/blob/main/projects_aug_2…
-
Andrew Ng
@AndrewYNg
AI Dev 25 is coming to NYC on November 14!
1,200+ developers will dive into technical topics such as:
- Agentic AI: Multi-agent orchestration, tool use, complex reasoning chains
- Coding with AI: Agentic coding assistants, automated testing, debugging strategies
- Context engineering: Advanced RAG, structured context, memory systems
- Multimodal AI: Vision-language models, audio processing, cross-modal architectures
- Fintech applications: Fraud detection, credit modeling, regulatory compliance
Our Pi Day AI Dev event sold out quickly, so we booked a bigger venue this time. Tickets available here: ai-dev.deeplearning.ai -
Andrew Ng
@AndrewYNg
Just as many businesses are transforming to become more capable by using AI, universities are too. I recently visited the UK to receive an honorary doctorate from the University of Exeter’s Faculty of Environment, Science and Economy. @UniofExeter The name of this faculty stood out to me as a particularly forward-looking way to organize an academic division. Having Computer Science sit alongside Environmental Science and the Business School creates natural opportunities for collaboration across these fields.
Leveraging AI leads a university to do things differently. Speaking with Vice Chancellor Lisa Roberts, Deputy Vice Chancellor Timothy Quine, and CS Department Head Andrew Howes, I was struck by the university leadership’s pragmatic and enthusiastic embrace of AI. This is not a group whose primary worry is whether students will cheat using AI. This is a group that is thinking about how to create a student body that is empowered through AI, whether by teaching more students to code, helping them use AI tools effectively, or showing them what’s newly possible in their disciplines.
Exeter is a wonderful place to create synergies between AI, environmental science, and business. It hosts 5 of the world’s top 21 most influential climate scientists according to Reuters, and its scholars are major contributors to reports by the UN’s IPCC (Intergovernmental Panel on Climate Change) as well as pioneers in numerous areas of climate research including geoengineering, which I wrote about previously. Its Centre for Environmental Intelligence, a partnership with the Met Office (the UK’s national weather service), applies AI to massive climate datasets. More work like this is needed to understand climate change and strategies for mitigation and adaptation. Add to this its Business School — named Business School of the Year by the consultancy Times Higher Education — and you have the ingredients for building applications and pursuing interdisciplinary studies that span technological, environmental, and economic realities.
Having been born in the UK and spent most of my career in Silicon Valley, I find it exciting to see Exeter’s leadership embrace AI with an enthusiasm I more often associate with California. The UK has always punched above its weight in research, and seeing that tradition continue in the AI era is encouraging.
Just as every company is becoming an AI company, every university must become an AI university — not just teaching AI, but using it to advance every field of study. This doesn’t mean abandoning disciplinary expertise. It means maintaining technical excellence while ensuring AI enhances every field.
Like almost all other universities and businesses worldwide, Exeter’s AI transformation is just beginning. But the enthusiastic embrace of AI by its leadership will give it momentum. As someone who is proud to be an honorary graduate of the university, I look forward to seeing what comes next!
[Original text: www.deeplearning.ai/the-batch/issue-314/ ] -
Andrew Ng
@AndrewYNg
Buildathon: The Rapid Engineering Competition livestreams this Saturday, August 16. Top developers will compete to build 5+ products in a single day using AI coding assistants – projects that traditionally took weeks. Watch live as they advance through semifinals and finals, and see how fast software can now be built! Register at buildathon.ai
-
Andrew Ng
@AndrewYNg
Recently Meta made headlines with unprecedented, massive compensation packages for AI model builders exceeding $100M (sometimes spread over multiple years). With the company planning to spend $66B-72B this year on capital expenses such as data centers, a meaningful fraction of which will be devoted to AI, from a purely financial point of view, it’s not irrational to spend a few extra billion dollars on salaries to make sure this hardware is used well.
A typical software-application startup that’s not involved in training foundation models might spend 70-80% of its dollars on salaries, 5-10% on rent, and 10-25% on other operating expenses (cloud hosting, software licenses, marketing, legal/accounting, etc.). But scaling up models is so capital-intensive, salaries are a small fraction of the overall expense. This makes it feasible for businesses in this area to pay their relatively few employees exceptionally well. If you’re spending tens of billions of dollars on GPU hardware, why not spend just a tenth of that on salaries? Even before Meta’s recent offers, salaries of AI model trainers have been high, with many being paid $5-10M/year, although Meta has raised these numbers to new heights.
Meta carries out many activities, including run Facebook, Instagram, WhatsApp, and Oculus. But the Llama/AI-training part of its operations is particularly capital-intensive. Many of Meta’s properties rely on user-generated content (UGC) to attract attention, which is then monetized through advertising. AI is a huge threat and opportunity to such businesses: If AI-generated content (AIGC) substitutes for UGC to capture people's attention to sell ads against, this will transform the social-media landscape.
This is why Meta — like TikTok, YouTube, and other social-media properties — is paying close attention to AIGC, and why making significant investments in AI is rational. Further, when Meta hires a key employee, not only does it gain the future work output of that person, but it also potentially gets insight into a competitor’s technology, which also makes its willingness to pay high salaries a rational business move (so long as it does not adversely affect the company’s culture).
The pattern of capital-intensive businesses compensating employees extraordinarily well is not new. For example, Netflix expects to spend a huge $18B this year on content. This makes the salary expense of paying its 14,000 employees a small fraction of the total expense, which allows the company to routinely pay above-market salaries. Its ability to spend this way also shapes a distinctive culture that includes elements of “we’re a sports team, not a family” (which seems to work for Netflix but isn’t right for everyone). In contrast, a labor-intensive manufacturing business like Foxconn, which employs over 1 million people globally, has to be much more price-sensitive in what it pays people.
Even a decade ago, when I led a team that worked to scale up AI, I built spreadsheets that modeled how much of my budget to allocate toward salaries and how much to allocate toward GPUs (using a custom model for how much productive output N employees and M GPUs would lead to, so I could optimize N and M subject to my budget constraint). Since then, the business of scaling up AI has skewed the spending significantly toward GPUs.
I’m happy for the individuals who are getting large pay packages. And regardless of any individual's pay, I’m grateful for the contributions of everyone working in AI. Everyone in AI deserves a good salary, and while the gaps in compensation are growing, I believe this reflects the broader phenomenon that developers who work in AI, at this moment in history, have an opportunity to make a huge impact and do world-changing work.
[Original text: www.deeplearning.ai/the-batch/issue-313/ ] -
Andrew Ng
@AndrewYNg
I'm thrilled to announce the definitive course on Claude Code, created with @AnthropicAI and taught by Elie Schoppik @eschoppik. If you want to use highly agentic coding - where AI works autonomously for many minutes or longer, not just completing code snippets - this is it.
Claude Code has been a game-changer for many developers (including me!), but there's real depth to using it well. This comprehensive course covers everything from fundamentals to advanced patterns.
After this short course, you'll be able to:
- Orchestrate multiple Claude subagents to work on different parts of your codebase simultaneously
- Tag Claude in GitHub issues and have it autonomously create, review, and merge pull requests
- Transform messy Jupyter notebooks into clean, production-ready dashboards
- Use MCP tools like Playwright so Claude can see what's wrong with your UI and fix it autonomously
Whether you're new to Claude Code or already using it, you'll discover powerful capabilities that can fundamentally change how you build software.
I'm very excited about what agentic coding lets everyone now do. Please take this course!
www.deeplearning.ai/short-courses/claude-code-a-highly-ag…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I'm thrilled @OpenAI has released two open weight models. Thank you to all my friends at OpenAI for this gift! I'm also encouraged that from my quick tests gpt-oss-120b looks strong (though we should still wait for rigorous 3rd party evals).
-
Andrew Ng
@AndrewYNg
There is now a path for China to surpass the U.S. in AI. Even though the U.S. is still ahead, China has tremendous momentum with its vibrant open-weights model ecosystem and aggressive moves in semiconductor design and manufacturing. In the startup world, we know momentum matters: Even if a company is small today, a high rate of growth compounded for a few years quickly becomes an unstoppable force. This is why a small, scrappy team with high growth can threaten even behemoths. While both the U.S. and China are behemoths, China’s hypercompetitive business landscape and rapid diffusion of knowledge give it tremendous momentum. The White House’s AI Action Plan released last week, which explicitly champions open source (among other things), is a very positive step for the U.S., but by itself it won’t be sufficient to sustain the U.S. lead.
Now, AI isn’t a single, monolithic technology, and different countries are ahead in different areas. For example, even before Generative AI, the U.S. had long been ahead in scaled cloud AI implementations, while China has long been ahead in surveillance technology. These translate to different advantages in economic growth as well as both soft and hard power. Even though nontechnical pundits talk about “the race to AGI” as if AGI were a discrete technology to be invented, the reality is that AI technology will progress continuously, and there is no single finish line. If a company or nation declares that it has achieved AGI, I expect that declaration to be less a technology milestone than a marketing milestone. A slight speed advantage in the Olympic 100m dash translates to a dramatic difference between winning a gold medal versus a silver medal. An advantage in AI prowess translates into a proportionate advantage in economic growth and national power; while the impact won’t be a binary one of either winning or losing everything, these advantages nonetheless matter.
Looking at Artificial Analysis and LMArena leaderboards, the top proprietary models were developed in the U.S., but the top open models come from China. Google’s Gemini 2.5 Pro, OpenAI’s o4, Anthropic’s Claude 4 Opus, and Grok 4 are all strong models. But open alternatives from China such as DeepSeek R1-0528, Kimi K2 (designed for agentic reasoning), Qwen3 variations (including Qwen3-Coder, which is strong at coding) and Zhipu’s GLM 4.5 (whose post-training software was released as open source) are close behind, and many are ahead of Meta’s Llama 4 and Google’s Gemma 3 — the U.S.’ best open-weights offerings.
Because many U.S. companies have taken a secretive approach to developing foundation models — a reasonable business strategy — the leading companies spend huge numbers of dollars to recruit key team members from each other who might know the “secret sauce“ that enabled a competitor to develop certain capabilities. So knowledge does circulate, but at high cost and slowly. In contrast, in China’s open AI ecosystem, many advanced foundation model companies undercut each other on pricing, make bold PR announcements, and poach each others’ employees and customers. This Darwinian life-or-death struggle will lead to the demise of many of the existing players, but the intense competition breeds strong companies.
In semiconductors, too, China is making progress. Huawei’s CloudMatrix 384 aims to compete with Nvidia’s GB200 high-performance computing system. While China has struggled to develop GPUs with a similar capability as Nvidia’s top-of-the-line B200, Huawei is trying to build a competitive system by combining a larger number (384 instead of 72) of lower-capability chips. China’s automotive sector once struggled to compete with U.S. and European internal combustion engine vehicles, but leapfrogged ahead by betting on electric vehicles. It remains to be seen how effective Huawei’s alternative architectures prove to be, but the U.S. export restrictions have given Huawei and other Chinese businesses a strong incentive to invest heavily in developing their own technology. Further, if China were to develop its domestic semiconductor manufacturing capabilities while the U.S. remained reliant on TSMC in Taiwan, then the U.S.’ AI roadmap would be much more vulnerable to a disruption of the Taiwan supply chain (perhaps due to a blockade or, worse, a hot war).
With the rise of electricity, the internet, and other general-purpose technologies, there was room for many nations to benefit, and the benefit to one nation hasn’t come at the expense of another. I know of businesses that, many months back, planned for a future in which China dominates open models (indeed, we are there at this moment, although the future depends on our actions). Given the transformative impact of AI, I hope all nations — especially democracies with a strong respect for human rights and the rule of law — will clear roadblocks from AI progress and invest in open science and technology to increase the odds that this technology will support democracy and benefit the greatest possible number of people.
[Full text: www.deeplearning.ai/the-batch/issue-312/ ] -
Andrew Ng reposted
Andrea Azzini
@andreaazzini
This was a somewhat contrarian take 6 months ago. Now it’s something I see most people agreeing with. The best PMs have gained months of momentum
Andrew Ng
@AndrewYNg
Writing software, especially prototypes, is becoming cheaper. This will lead to increased demand for people who can decide what to build. AI Product Management has a bright future!
Software is often written by teams that comprise Product Managers (PMs), who decide what to build (such as what features to implement for what users) and Software Developers, who write the code to build the product. Economics shows that when two goods are complements — such as cars (with internal-combustion engines) and gasoline — falling prices in one leads to higher demand for the other. For example, as cars became cheaper, more people bought them, which led to increased demand for gas. Something similar will happen in software. Given a clear specification for what to build, AI is making the building itself much faster and cheaper. This will significantly increase demand for people who can come up with clear specs for valuable things to build.
This is why I’m excited about the future of Product Management, the discipline of developing and managing software products. I’m especially excited about the future of AI Product Management, the discipline of developing and managing AI software products.
Many companies have an Engineer:PM ratio of, say, 6:1. (The ratio varies widely by company and industry, and anywhere from 4:1 to 10:1 is typical.) As coding becomes more efficient, teams will need more product management work (as well as design work) as a fraction of the total workforce. Perhaps engineers will step in to do some of this work, but if it remains the purview of specialized Product Managers, then the demand for these roles will grow.
This change in the composition of software development teams is not yet moving forward at full speed. One major force slowing this shift, particularly in AI Product Management, is that Software Engineers, being technical, are understanding and embracing AI much faster than Product Managers. Even today, most companies have difficulty finding people who know how to develop products and also understand AI, and I expect this shortage to grow.
Further, AI Product Management requires a different set of skills than traditional software Product Management. It requires:
- Technical proficiency in AI. PMs need to understand what products might be technically feasible to build. They also need to understand the lifecycle of AI projects, such as data collection, building, then monitoring, and maintenance of AI models.
- Iterative development. Because AI development is much more iterative than traditional software and requires more course corrections along the way, PMs need be able to manage such a process.
- Data proficiency. AI products often learn from data, and they can be designed to generate richer forms of data than traditional software.
- Skill in managing ambiguity. Because AI’s performance is hard to predict in advance, PMs need to be comfortable with this and have tactics to manage it.
- Ongoing learning. AI technology is advancing rapidly. PMs, like everyone else who aims to make best use of the technology, need to keep up with the latest technology advances, product ideas, and how they fit into users’ lives.
Finally, AI Product Managers will need to know how to ensure that AI is implemented responsibly (for example, when we need to implement guardrails to prevent bad outcomes), and also be skilled at gathering feedback fast to keep projects moving. Increasingly, I also expect strong product managers to be able to build prototypes for themselves.
The demand for good AI Product Managers will be huge. In addition to growing AI Product Management as a discipline, perhaps some engineers will also end up doing more product management work.
The variety of valuable things we can build is nearly unlimited. What a great time to build!
[Original text: www.deeplearning.ai/the-batch/issue-284/ ] -
Andrew Ng
@AndrewYNg
Announcing our new event - Buildathon: The Rapid Engineering Competition. See the video for details, and please apply to participate!
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
The invention of modern writing instruments like the typewriter made writing easier, but they also led to the rise of writer’s block, where deciding what to write became the bottleneck. Similarly, the invention of agentic coding assistants has led to a new builder’s block, where the holdup is deciding what to build. I call this the Product Management Bottleneck.
Product management is the art and science of deciding what to build. Because highly agentic coding accelerates the writing of software to a given product specification, deciding what to build is the new bottleneck, especially in early-stage projects. As the teams I work with take advantage of agentic coders, I increasingly value product managers (PMs) who have very high user empathy and can make product decisions quickly, so the speed of product decision-making matches the speed of coding.
PMs with high user empathy can make decisions by gut and get them right a lot of the time. As new information comes in, they can keep refining their mental models of what users like or do not like — and thereby refine their gut — and keep making fast decisions of increasing quality.
Many tactics are available to get user feedback and other forms of data that shape our beliefs about users. They include conversations with a handful of users, focus groups, surveys, and A/B tests on scaled products. But to drive progress at GenAI speed, I find that synthesizing all these sources of data in a PM's gut helps us move faster.
Let me illustrate with an example. Recently, my team debated which of 4 features users would prefer. I had my instincts, but none of us were sure, so we surveyed about 1,000 users. The results contradicted my initial beliefs — I was wrong! So what was the right thing to do at this point?
- Option 1: Go by the survey and build what users told us clearly they prefer.
- Option 2: Examine the survey data in detail to see how it changes my beliefs about what users want. That is, refine my mental model of users. Then use my revised mental model to decide what to do.
Even though some would consider Option 1 the “data-driven” way to make decisions, I consider this an inferior approach for most projects. Surveys may be flawed. Further, taking time to run a survey before making a decision results in slow decision-making.
In contrast, using Option 2, the survey results give much more generalizable information that can help me shape not just this decision, but many others as well. And it lets me process this one piece of data alongside all the user conversations, surveys, market reports, and observations of user behavior when they’re engaging with our product to form a much fuller view on how to serve users. Ultimately, that mental model drives my product decisions.
Of course, this technique does not always scale. For example, with programmatic online advertising in which AI might try to optimize the number of clicks on ads shown, an automated system conducts far more experiments in parallel and gathers data on what users do and do not click on, to filter through a PM's mental model of users. When a system needs to make a huge number of decisions, such as what ads to show (or products to recommend) on a huge number of pages, PM review and human intuition do not scale.
But in products where a team is making a small number of critical decisions such as what key features to prioritize, I find that data — used to help build a good mental model of the user, which is then applied to make decisions very quickly — is still the best way to drive rapid progress and relieve the Product Management Bottleneck.
[Original text: www.deeplearning.ai/the-batch/issue-310/ ] -
Andrew Ng
@AndrewYNg
Announcing a new Coursera course: Retrieval Augmented Generation (RAG)
You'll learn to build high performance, production-ready RAG systems in this hands-on, in-depth course created by deeplearning.ai and taught by @zainhas, experienced AI and ML engineer, researcher, and educator.
RAG is a critical component today of many LLM-based applications in customer support, internal company Q&A systems, even many of the leading chatbots that use web search to answer your questions. This course teaches you in-depth how to make RAG work well.
LLMs can produce generic or outdated responses, especially when asked specialized questions not covered in its training data. RAG is the most widely used technique for addressing this. It brings in data from new data sources, such as internal documents or recent news, to give the LLM the relevant context to private, recent, or specialized information. This lets it generate more grounded and accurate responses.
In this course, you’ll learn to design and implement every part of a RAG system, from retrievers to vector databases to generation to evals. You’ll learn about the fundamental principles behind RAG and how to optimize it at both the component and whole-system levels.
As AI evolves, RAG is evolving too. New models can handle longer context windows, reason more effectively, and can be parts of complex agentic workflows. One exciting growth area is Agentic RAG, in which an AI agent at runtime (rather than it being hardcoded at development time) autonomously decides what data to retrieve, and when/how to go deeper. Even with this evolution, access to high-quality data at runtime is essential, which is why RAG is a key part of so many applications.
You'll learn via hands-on experiences to:
- Build a RAG system with retrieval and prompt augmentation
- Compare retrieval methods like BM25, semantic search, and Reciprocal Rank Fusion
- Chunk, index, and retrieve documents using a Weaviate vector database and a news dataset
- Develop a chatbot, using open-source LLMs hosted by Together AI, for a fictional store that answers product and FAQ questions
- Use evals to drive improving reliability, and incorporate multi-modal data
RAG is an important foundational technique. Become good at it through this course!
Please sign up here: www.coursera.org/learn/retrieval-augmented-generation-rag -
Andrew Ng
@AndrewYNg
Announcing AI Aspire, our new advisory firm to help enterprises with their AI strategy and transformation journey! We are partnering with Bain & Company and looking forward to helping businesses unlock scalable, transformative value.
C-suite is now realizing that top-down leadership is needed for AI transformation. But figuring out the implications of AI on a particular business is extremely complex. The Bain team is world class at helping businesses craft strategy and navigate complex landscapes. Kirsty Tan (at AI Aspire) and I are thrilled to be working with Bain's Chuck Whitten, Sarah Elk, Erika Serow and team to help businesses.
Questions that C-suite and boards are now asking include:
- What new products are now possible? What can we (or our competitors) now do for customers that were not possible before?
- How do we use AI to boost productivity? What workflows can be streamlined or processes reinvented?
- What technology investments, for example in data infrastructure, should now be prioritized? And what are the risks, for example to security and compliance?
- What are the HR implications: What roles do we hire more or fewer of?
- How do we bring the team with us for the journey, and enable transformation from within?
- Does AI enable new competitors; or alternatively, are there now new markets that are possible for us to move into?
Technology is no longer merely a support system for business — it is engine of growth for the companies nimble enough to adapt and daring enough to lead.
Press release: www.aiaspire.ai/press-release-1/bain -
Andrew Ng
@AndrewYNg
Last week, the United States Congress passed President Trump’s “Big Beautiful Bill.” I’m disappointed it didn’t include a proposed moratorium on U.S. state-level AI regulation. While there is a role for AI regulation, it is when the technology is new and poorly understood that lobbyists are most likely to succeed at pushing through anti-competitive regulations that hamper open-source and other beneficial AI efforts. A moratorium would have bought more time for regulators to figure out the realistic risks and rewards of AI and thereby avoid bad regulatory proposals.
Many jurisdictions loosely follow this trajectory:
- When new AI technology is still poorly understood, companies can make grandiose statements about its benefits or dangers, and both traditional and social media are ineffective at fact-checking them and tend to parrot what they say. During this initial period, businesses can get away with saying almost anything.
- This opens opportunities for hype as well as fear mongering based on exaggerated claims about AI dangers. Some businesses exploit this opportunity to try to get regulators to pass anti-competitive laws that impede open-source and other competitors.
- But eventually, smart regulators learn enough about AI to understand its realistic benefits and risks. For example, the U.S. Senate’s bipartisan Insight Forum on AI, which I participated in, heard from many stakeholders and came to support innovation and dismiss ill-founded fears of “AI takeover” and the like.
Indeed, the European Union went through this trajectory as well. After the EU AI Act was passed, many regulators realized many of its “protections” are not actually helpful. They relaxed some of the law’s provisions to make it less stifling of innovation than many observers initially had feared.
There are AI regulations that would limit harmful applications appropriately, for example, banning non-consensual deepfake porn and preventing misleading marketing. However, many states, which have less resources than the federal government to deeply understand AI, have proposed harmful regulations, especially those that aim to regulate the technology rather than the applications.
For example:
- California’s SB 1047 purported to impose safety requirements on frontier AI systems, but it placed ambiguous and/or technically infeasible requirements on model creators to prevent harmful downstream uses. This is akin to holding the maker of a hammer liable if someone uses it for harmful purposes. Fortunately, Governor Gavin Newsom quashed SB 1047 with a veto.
- New York’s Responsible AI Safety and Education Act, which passed the state legislature in June and awaits Governor Kathy Hochul’s signature or veto, also places ambiguous and unreasonable requirements on model builders, purportedly to guard against theoretical “critical harms.” It would hamper open source without making anyone meaningfully safer.
- The Texas Responsible AI Governance Act initially included many of the problematic elements of SB 1047. It would have created unreasonable requirements that model providers would have had a hard time complying with, and compliance would have amounted to safety theater that was unlikely to actually make people safer. Fortunately, as Texas regulators came to understand AI better, they significantly scaled back the law, and Governor Greg Abbott signed it into law in late June. The final law focuses on specific application areas, establishes an advisory council and regulatory sandbox, and places more burden on government agencies than private companies.
Sadly, I see the net impact of the regulations proposed so far as negative. Many would severely hamper innovation despite some lesser positive benefits. This is why a moratorium on state-level regulation would have been a net benefit to AI and to society. Shutting down bad regulations for a limited period would have given regulators time to figure out AI technology and ignore irresponsible fear mongering. In addition, it would have helped them avoid creating a patchwork of state-level regulations that businesses large and small have a hard time complying with.
Perhaps a 10-year blanket moratorium was a step too far. A more modest, say, 2-year moratorium, and one that covered only the most problematic regulatory proposals, might have had a better chance of passing.
Even though a moratorium did not make it into Trump’s bill, I hope that efforts continue in the U.S. and other nations to give regulators time to understand the real risks and benefits of AI, and not pass stifling regulations during that initial period when the technology is new and the power of fear mongering is strongest.
[Original text: www.deeplearning.ai/the-batch/issue-309/ ] -
Andrew Ng
@AndrewYNg
My talk at YC Startup School on how to build AI startups. I share tips from @AI_Fund on how to use AI to build fast. Let me know what you think!
Y Combinator
@ycombinator
Andrew Ng (@AndrewYNg) on how startups can build faster with AI.
At AI Startup School in San Francisco.
00:31 - The Importance of Speed in Startups
01:13 - Opportunities in the AI Stack
02:06 - The Rise of Agent AI
04:52 - Concrete Ideas for Faster Execution
08:56 - Rapid Prototyping and Engineering
17:06 - The Role of Product Management
21:23 - The Value of Understanding AI
22:33 - Technical Decisions in AI Development
23:26 - Leveraging Gen AI Tools for Startups
24:05 - Building with AI Building Blocks
25:26 - The Importance of Speed in Startups
26:41 - Addressing AI Hype and Misconceptions
37:35 - AI in Education: Current Trends and Future Directions
39:33 - Balancing AI Innovation with Ethical Considerations
41:27 - Protecting Open Source and the Future of AIThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Agentic Document Extraction now supports field extraction! Many doc extraction use cases extract specific fields from forms and other structured documents. You can now input a picture or PDF of an invoice, request the vendor name, item list, and prices, and get back the extracted fields. Or input a medical form and specify a schema to extract patient name, patient ID, insurance number, etc.
One cool feature: If you don't feel like writing a schema (json specification of what fields to extract) yourself, upload one sample document and write a natural language prompt saying what you want, and we automatically generate a schema for you.
See the video for details!This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New Course: Post-training of LLMs
Learn to post-train and customize an LLM in this short course, taught by @BanghuaZ, Assistant Professor at the University of Washington @UW, and co-founder of @NexusflowX.
Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning.
Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training.
In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance.
You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior.
In detail, you’ll:
- Understand what post-training is, when to use it, and how it differs from pre-training.
- Build an SFT pipeline to turn a base model into an instruct model.
- Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones.
- Implement a DPO pipeline to change the identity of a chat assistant.
- Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions.
- Train a model with GRPO to improve its math capabilities using a verifiable reward.
Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today.
Please sign up here: www.deeplearning.ai/short-courses/post-training-of-llms/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I’d like to share a tip for getting more practice building with AI — that is, either using AI building blocks to build applications or using AI coding assistance to create powerful applications quickly: If you find yourself with only limited time to build, reduce the scope of your project until you can build something in whatever time you do have.
If you have only an hour, find a small component of an idea that you're excited about that you can build in an hour. With modern coding assistants like Anthropic’s Claude Code (my favorite dev tool right now), you might be surprised at how much you can do even in short periods of time! This gets you going, and you can always continue the project later.
To become good at building with AI, most people must (i) learn relevant techniques, for example by taking online AI courses, and (ii) practice building. I know developers who noodle on ideas for months without actually building anything — I’ve done this too! — because we feel we don’t have time to get started. If you find yourself in this position, I encourage you to keep cutting the initial project scope until you identify a small component you can build right away.
Let me illustrate with an example — one of my many small, fun weekend projects that might never go anywhere, but that I’m glad I did.
Here’s the idea: Many people fear public speaking. And public speaking is challenging to practice, because it's hard to organize an audience. So I thought it would be interesting to build an audience simulator to provide a digital audience of dozens to hundreds of virtual people on a computer monitor and let a user practice by speaking to them.
One Saturday afternoon, I found myself in a coffee shop with a couple of hours to spare and decided to give the audience simulator a shot. My familiarity with graphics coding is limited, so instead of building a complex simulator of a large audience and writing AI software to simulate appropriate audience responses, I decided to cut scope significantly to (a) simulating an audience of one person (which I could replicate later to simulate N persons), (b) omitting AI and letting a human operator manually select the reaction of the simulated audience (similar to Wizard of Oz prototyping), and (c) implementing the graphics using a simple 2D avatar.
Using a mix of several coding assistants, I built a basic version in the time I had. The avatar could move subtly and blink, but otherwise it used basic graphics. Even though it fell far short of a sophisticated audience simulator, I am glad I built this. In addition to moving the project forward and letting me explore different designs, it advanced my knowledge of basic graphics. Further, having this crude prototype to show friends helped me get user feedback that shaped my views on the product idea.
I have on my laptop a list of ideas of things that I think would be interesting to build. Most of them would take much longer than the handful of hours I might have to try something on a given day, but by cutting their scope, I can get going, and the initial progress on a project helps me decide if it’s worth further investment. As a bonus, hacking on a wide variety of applications helps me practice a wide range of skills. But most importantly, this gets an idea out of my head and potentially in front of prospective users for feedback that lets the project move faster.
[Original text: www.deeplearning.ai/the-batch/issue-308/ ] -
Andrew Ng
@AndrewYNg
On Monday, a United States District Court ruled that training LLMs on copyrighted books constitutes fair use. A number of authors had filed suit against Anthropic for training its models on their books without permission. Just as we allow people to read books and learn from them to become better writers, but not to regurgitate copyrighted text verbatim, the judge concluded that it is fair use for AI models to do so as well.
Indeed, Judge Alsup wrote that the authors’ lawsuit is “no different than it would be if they complained that training schoolchildren to write well would result in an explosion of competing works.” While it remains to be seen whether the decision will be appealed, this ruling is reasonable and will be good for AI progress. (Usual caveat: I am not a lawyer and am not giving legal advice.)
AI has massive momentum, but a few things could put progress at risk:
- Regulatory capture that stifles innovation, including especially open source
- Loss of access to cutting-edge semiconductor chips (the most likely cause would be war breaking out in Taiwan)
- Regulations that severely impede access to data for training AI systems
Access to high-quality data is important. Even though the mass media tends to talk about the importance of building large data centers and scaling up models, when I speak with friends at companies that train foundation models, many describe a very large amount of their daily challenges as data preparation. Specifically, a significant fraction of their day-to-day work follows the usual Data Centric AI practices of identifying high-quality data (books are one important source), cleaning data (the ruling describes Anthropic taking steps like removing book pages' headers, footers, and page numbers), carrying out error analyses to figure out what types of data to acquire more of, and inventing new ways to generate synthetic data.
I am glad that a major risk to data access just decreased. Appropriately, the ruling further said that Anthropic’s conversion of books from paper format to digital — a step that’s needed to enable training — also was fair use. However, in a loss for Anthropic, the judge indicated that, while training on data that was acquired legitimately is fine, using pirated materials (such as texts downloaded from pirate websites) is not fair use. Thus, Anthropic still may be liable on this point. Other LLM providers, too, will now likely have to revisit their practices if they use datasets that may contain pirated works.
Overall, the ruling is positive for AI progress. Perhaps the biggest benefit is that it reduces ambiguity with respect to AI training and copyright and (if it stands up to appeals) makes the roadmap for compliance clearer. This decision indicates it is okay to train on legitimately acquired data to build models that generate transformational outputs, and to convert printed books to digital format for this purpose. However, downloading from pirate sites (as well as permanently building a “general purpose” library of texts, stored indefinitely for purposes to be determined, without permission from the relevant copyright holders) are not considered fair use.
I am very sympathetic with the many writers who are worried about their livelihoods being affected by AI. I don‘t know the right solution for that. Society is better off with free access to more data; but if a subset of people is significantly negatively affected, I hope we can figure out an arrangement that compensates them fairly.
[Original text: www.deeplearning.ai/the-batch/issue-307/ ] -
Andrew Ng
@AndrewYNg
New Course: ACP: Agent Communication Protocol
Learn to build agents that communicate and collaborate across different frameworks using ACP in this short course built with @IBMResearch's BeeAI, and taught by @sandi_besen, AI Research Engineer & Ecosystem Lead at IBM, and @nicholasrenotte, Head of AI Developer Advocacy at IBM.
Building a multi-agent system with agents built or used by different teams and organizations can become challenging. You may need to write custom integrations each time a team updates their agent design or changes their choice of agentic orchestration framework.
The Agent Communication Protocol (ACP) is an open protocol that addresses this challenge by standardizing how agents communicate, using a unified RESTful interface that works across frameworks. In this protocol, you host an agent inside an ACP server, which handles requests from an ACP client and passes them to the appropriate agent. Using a standardized client-server interface allows multiple teams to reuse agents across projects. It also makes it easier to switch between frameworks, replace an agent with a new version, or update a multi-agent system without refactoring the entire system.
In this course, you’ll learn to connect agents through ACP. You’ll understand the lifecycle of an ACP Agent and how it compares to other protocols, such as MCP (Model Context Protocol) and A2A (Agent-to-Agent). You’ll build ACP-compliant agents and implement both sequential and hierarchical workflows of multiple agents collaborating using ACP.
Through hands-on exercises, you’ll build:
- A RAG agent with CrewAI and wrap it inside an ACP server.
- An ACP Client to make calls to the ACP server you created.
- A sequential workflow that chains an ACP server, created with Smolagents, to the RAG agent.
- A hierarchical workflow using a router agent that transforms user queries into tasks, delegated to agents available through ACP servers.
- An agent that uses MCP to access tools and ACP to communicate with other agents.
You’ll finish up by importing your ACP agents into the BeeAI platform, an open-source registry for discovering and sharing agents.
ACP enables collaboration between agents across teams and organizations. By the end of this course, you’ll be able to build ACP agents and workflows that communicate and collaborate regardless of framework.
Please sign up here: www.deeplearning.ai/short-courses/acp-agent-communication…This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Erik Brynjolfsson
@erikbryn
As an immigrant myself and as a person who has had the privilege of working closely with hundreds of brilliant immigrants, I join @AndrewYNg in supporting high-skilled, legal immigration.
It's America's greatest competitive advantage.
It's tragic to see it being destroyed.
Andrew Ng
@AndrewYNg
One of the most effective things the U.S. or any other nation can do to ensure its competitiveness in AI is to welcome high-skilled immigration and international students who have the potential to become high-skilled. For centuries, the U.S. has welcomed immigrants, and this helped make it a worldwide leader in technology. Letting immigrants and native-born Americans collaborate makes everyone better off. Reversing this stance would have a huge negative impact on U.S. technology development.
I was born in the UK and came to the U.S. on an F-1 student visa as a relatively unskilled and clueless teenager to attend college. Fortunately I gained skills and became less clueless over time. After completing my graduate studies, I started working at Stanford under the OPT (Optional Practical Training) program, and later an H-1B visa, and ended up staying here. Many other immigrants have followed similar paths to contribute to the U.S.
I am very concerned that making visas harder to obtain for students and high-skilled workers, such as the pause in new visa interviews that started last month and a newly chaotic process of visa cancellations, will hurt our ability to attract great students and workers. In addition, many international students without substantial means count on being able to work under OPT to pay off the high cost of a U.S. college degree. Gutting the OPT program, as has been proposed, would both hurt many international students’ ability to study here and deprive U.S. businesses of great talent. (This won’t stop students from wealthy families. But the U.S. should try to attract the best talent without regard to wealth.)
Failure to attract promising students and high-skilled workers would have a huge negative impact on American competitiveness in AI. Indeed, a recent report by the National Security Commission on Artificial Intelligence exhorts the government to “strengthen AI talent through immigration.”
If talented people do not come to the U.S., will they have an equal impact on global AI development just working somewhere else? Unfortunately, the net impact will be negative. The U.S. has a number of tech hubs including Silicon Valley, Seattle, New York, Boston/Cambridge, Los Angeles, Pittsburgh and Austin, and these hubs concentrate talent and foster innovation. (This is why cities, where people can more easily find each other and collaborate, promote innovation.) Making it harder for AI talent to find each other and collaborate will slow down innovation, and it will take time for new hubs to become as advanced.
Nonetheless, other nations are working hard to attract immigrants who can drive innovation — a good move for them! Many have thoughtful programs to attract AI and other talent. There are the UK’s Global Talent Visa, France’s French Tech Visa, Australia’s Global Talent Visa, the UAE’s Golden Visa, Taiwan’s Employment Gold Card, China’s Thousand Talents Plan, and many more. The U.S. is fortunate that many people already want to come here to study and work. Squandering that advantage would be a huge unforced error.
Beyond the matter of national competitiveness, there is the even more important ethical matter of making sure people are treated decently. I have spoken with international students who are terrified that their visas may be canceled arbitrarily. One recently agonized about whether to attend an international conference to present a research paper, because they were worried about being unable to return. In the end, with great sadness, they cancelled their trip. I also spoke with a highly skilled technologist who is in the U.S. on an H-1B visa. Their company shut down, leading them — after over a decade in this country, and with few ties to their nation of origin — scrambling to find alternative employment that would enable them to stay.
These stories, and many far worse, are heartbreaking. While I do what I can to help individuals I know personally, it is tragic that we are creating such an uncertain environment for immigrants, that many people who have extraordinary skills and talents will no longer want to come here.
To every immigrant or migrant in the U.S. who is concerned about the current national environment: I see you and empathize with your worries. As an immigrant myself, I will be fighting to protect everyone’s dignity and right to due process, and to encourage legal immigration, which makes both the U.S. and individuals much better off.
[Full text, with links: www.deeplearning.ai/the-batch/issue-306/ ] -
Andrew Ng
@AndrewYNg
One of the most effective things the U.S. or any other nation can do to ensure its competitiveness in AI is to welcome high-skilled immigration and international students who have the potential to become high-skilled. For centuries, the U.S. has welcomed immigrants, and this helped make it a worldwide leader in technology. Letting immigrants and native-born Americans collaborate makes everyone better off. Reversing this stance would have a huge negative impact on U.S. technology development.
I was born in the UK and came to the U.S. on an F-1 student visa as a relatively unskilled and clueless teenager to attend college. Fortunately I gained skills and became less clueless over time. After completing my graduate studies, I started working at Stanford under the OPT (Optional Practical Training) program, and later an H-1B visa, and ended up staying here. Many other immigrants have followed similar paths to contribute to the U.S.
I am very concerned that making visas harder to obtain for students and high-skilled workers, such as the pause in new visa interviews that started last month and a newly chaotic process of visa cancellations, will hurt our ability to attract great students and workers. In addition, many international students without substantial means count on being able to work under OPT to pay off the high cost of a U.S. college degree. Gutting the OPT program, as has been proposed, would both hurt many international students’ ability to study here and deprive U.S. businesses of great talent. (This won’t stop students from wealthy families. But the U.S. should try to attract the best talent without regard to wealth.)
Failure to attract promising students and high-skilled workers would have a huge negative impact on American competitiveness in AI. Indeed, a recent report by the National Security Commission on Artificial Intelligence exhorts the government to “strengthen AI talent through immigration.”
If talented people do not come to the U.S., will they have an equal impact on global AI development just working somewhere else? Unfortunately, the net impact will be negative. The U.S. has a number of tech hubs including Silicon Valley, Seattle, New York, Boston/Cambridge, Los Angeles, Pittsburgh and Austin, and these hubs concentrate talent and foster innovation. (This is why cities, where people can more easily find each other and collaborate, promote innovation.) Making it harder for AI talent to find each other and collaborate will slow down innovation, and it will take time for new hubs to become as advanced.
Nonetheless, other nations are working hard to attract immigrants who can drive innovation — a good move for them! Many have thoughtful programs to attract AI and other talent. There are the UK’s Global Talent Visa, France’s French Tech Visa, Australia’s Global Talent Visa, the UAE’s Golden Visa, Taiwan’s Employment Gold Card, China’s Thousand Talents Plan, and many more. The U.S. is fortunate that many people already want to come here to study and work. Squandering that advantage would be a huge unforced error.
Beyond the matter of national competitiveness, there is the even more important ethical matter of making sure people are treated decently. I have spoken with international students who are terrified that their visas may be canceled arbitrarily. One recently agonized about whether to attend an international conference to present a research paper, because they were worried about being unable to return. In the end, with great sadness, they cancelled their trip. I also spoke with a highly skilled technologist who is in the U.S. on an H-1B visa. Their company shut down, leading them — after over a decade in this country, and with few ties to their nation of origin — scrambling to find alternative employment that would enable them to stay.
These stories, and many far worse, are heartbreaking. While I do what I can to help individuals I know personally, it is tragic that we are creating such an uncertain environment for immigrants, that many people who have extraordinary skills and talents will no longer want to come here.
To every immigrant or migrant in the U.S. who is concerned about the current national environment: I see you and empathize with your worries. As an immigrant myself, I will be fighting to protect everyone’s dignity and right to due process, and to encourage legal immigration, which makes both the U.S. and individuals much better off.
[Full text, with links: www.deeplearning.ai/the-batch/issue-306/ ] -
Andrew Ng
@AndrewYNg
Introducing "Building with Llama 4." This short course is created with @Meta @AIatMeta, and taught by @asangani7, Director of Partner Engineering for Meta’s AI team.
Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve.
In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis!
In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning.
If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context.
Please sign up here: www.deeplearning.ai/short-courses/building-with-llama-4This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
There’s a new breed of GenAI Application Engineers who can build more-powerful applications faster than was possible before, thanks to generative AI. Individuals who can play this role are highly sought-after by businesses, but the job description is still coming into focus. Let me describe their key skills, as well as the sorts of interview questions I use to identify them.
Skilled GenAI Application Engineers meet two primary criteria: (i) They are able to use the new AI building blocks to quickly build powerful applications. (ii) They are able to use AI assistance to carry out rapid engineering, building software systems in dramatically less time than was possible before. In addition, good product/design instincts are a significant bonus.
AI building blocks. If you own a lot of copies of only a single type of Lego brick, you might be able to build some basic structures. But if you own many types of bricks, you can combine them rapidly to form complex, functional structures. Software frameworks, SDKs, and other such tools are like that. If all you know is how to call a large language model (LLM) API, that's a great start. But if you have a broad range of building block types — such as prompting techniques, agentic frameworks, evals, guardrails, RAG, voice stack, async programming, data extraction, embeddings/vectorDBs, model fine tuning, graphDB usage with LLMs, agentic browser/computer use, MCP, reasoning models, and so on — then you can create much richer combinations of building blocks.
The number of powerful AI building blocks continues to grow rapidly. But as open-source contributors and businesses make more building blocks available, staying on top of what is available helps you keep on expanding what you can build. Even though new building blocks are created, many building blocks from 1 to 2 years ago (such as eval techniques or frameworks for using vectorDBs) are still very relevant today.
AI-assisted coding. AI-assisted coding tools enable developers to be far more productive, and such tools are advancing rapidly. Github Copilot, first announced in 2021 (and made widely available in 2022), pioneered modern code autocompletion. But shortly after, a new breed of AI-enabled IDEs such as Cursor and Windsurf offered much better code-QA and code generation. As LLMs improved, these AI-assisted coding tools that were built on them improved as well.
Now we have highly agentic coding assistants such as OpenAI’s Codex and Anthropic’s Claude Code (which I really enjoy using and find impressive in its ability to write code, test, and debug autonomously for many iterations). In the hands of skilled engineers — who don’t just “vibe code” but deeply understand AI and software architecture fundamentals and can steer a system toward a thoughtfully selected product goal — these tools make it possible to build software with unmatched speed and efficiency.
I find that AI-assisted coding techniques become obsolete much faster than AI building blocks, and techniques from 1 or 2 years ago are far from today's best practices. Part of the reason for this might be that, while AI builders might use dozens (hundreds?) of different building blocks, they aren’t likely to use dozens of different coding assistance tools at once, and so the forces of Darwinian competition are stronger among tools. Given the massive investments in this space by Anthropic, Google, OpenAI, and other players, I expect the frenetic pace of development to continue, but keeping up with the latest developments in AI-assisted coding tools will pay off, since each generation is much better than the last.
Bonus: Product skills. In some companies, engineers are expected to take pixel-perfect drawings of a product, specified in great detail, and write code to implement it. But if a product manager has to specify even the smallest detail, this slows down the team. The shortage of AI product managers exacerbates this problem. I see teams move much faster if GenAI Engineers also have some user empathy as well at basic skill at designing products, so that, given only high-level guidance on what to build (“a user interface that lets users see their profiles and change their passwords”), they can make a lot of decisions themselves and build at least a prototype to iterate from.
When interviewing GenAI Application Engineers, I will usually ask about their mastery of AI building blocks and ability to use AI-assisted coding, and sometimes also their product/design instincts. One additional question I've found highly predictive of their skill is, “How do you keep up with the latest developments in AI?” Because AI is evolving so rapidly, someone with good strategies for keeping up — such as reading The Batch and taking short courses 😃, regular hands-on practice building projects, and having a community to talk to — really does stay ahead of the game.
[Original post: www.deeplearning.ai/the-batch/issue-305/ ] -
Andrew Ng
@AndrewYNg
Learn to build and deploy GenAI pipelines in "Orchestrating Workflows for GenAI Applications", built in partnership with @astronomerio and taught by Kenten Danas, the company's DevRel Senior Manager, and Tamara Fingerlin, developer advocate.
Many GenAI applications require executing a pipeline comprising many steps. For example, a RAG app for recommending books might ingest and embed book descriptions, store the embeddings in a vector database, and later use the database to retrieve and recommend specific books based on a user query. After having prototyped this -- maybe in a Jupyter notebook -- how do you turn this into a reliable, repeatable workflow to run in production?
In this short course, you’ll learn to build reliable GenAI pipelines and orchestrate them using the popular open-source tool Airflow 3.0. You’ll learn to break down a workflow into discrete tasks so that an orchestration framework can schedule tasks to run in the right order at the right time (using time-based or data-aware triggers), and execute tasks in parallel when possible. It can also use retries to recover gracefully from failure (such as transient API rate limits) and provide observability (using Airflow UI) to help you track the status of the pipeline. You'll do this by using Airflow dags, which helps sequence tasks that need to run in a specific order, with clear task dependencies.
By the end of this course, you’ll know how to turn your prototype Jupyter notebook or Python script into production-ready workflow.
Please sign up here: www.deeplearning.ai/short-courses/orchestrating-workflows…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Hanging out with @juberti , OpenAI’s head of realtime AI, responsible for the company’s voice AI products. One thing both of us agree on: while some things in AI are overhyped, voice applications seem underhyped right now. The application opportunities seem larger than the amount of developer or business attention on this right now.
-
Andrew Ng
@AndrewYNg
Thank you for your pioneering research work @lateinteraction on DSPy (together with @matei_zaharia, @ChrisGPotts and many others). I'm glad we could do a short course on this!
Omar Khattab
@lateinteraction
Over the past year we received SO many requests for a @DeepLearningAI short course on DSPy.
Thanks to a great effort by my colleagues @ChenMoneyQ & @xqcathyyin from the DSPy team at @Databricks, here's a collaboration with @AndrewYNg's team who have been super helpful on this!
Andrew Ng
@AndrewYNg
New short course: DSPy: Build and Optimize Agentic Apps
DSPy is a powerful open-source framework for automatically tuning prompts for GenAI applications. In this course, you'll learn to use DSPy, together with MLflow. This is built in partnership with @databricks and taught by @ChenMoneyQ, co-lead of the DSPy framework.
Many AI builders spend hours hand-tuning prompts. When given a set of evals, DSPy automates this process. It’s especially useful for optimizing prompts, including few-shot prompts, in complex agentic AI workflows. Further, if you switch an application to a newer LLM, performance can degrade if your prompts were optimized to the previous model. DSPy automatically optimizes the entire system for the new LLM as well, using just a few evaluation examples.
This course teaches DSPy works, and best practices for using it. You’ll write programs using DSPy’s signature-based programming model, debug them with MLflow tracing -- to gain visibility into how different parts of a pipeline, as well as how the overall system, are performing -- and automatically improve their accuracy with DSPy Optimizer.
Please sign up here: www.deeplearning.ai/short-courses/dspy-build-optimize-age…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course: DSPy: Build and Optimize Agentic Apps
DSPy is a powerful open-source framework for automatically tuning prompts for GenAI applications. In this course, you'll learn to use DSPy, together with MLflow. This is built in partnership with @databricks and taught by @ChenMoneyQ, co-lead of the DSPy framework.
Many AI builders spend hours hand-tuning prompts. When given a set of evals, DSPy automates this process. It’s especially useful for optimizing prompts, including few-shot prompts, in complex agentic AI workflows. Further, if you switch an application to a newer LLM, performance can degrade if your prompts were optimized to the previous model. DSPy automatically optimizes the entire system for the new LLM as well, using just a few evaluation examples.
This course teaches DSPy works, and best practices for using it. You’ll write programs using DSPy’s signature-based programming model, debug them with MLflow tracing -- to gain visibility into how different parts of a pipeline, as well as how the overall system, are performing -- and automatically improve their accuracy with DSPy Optimizer.
Please sign up here: www.deeplearning.ai/short-courses/dspy-build-optimize-age…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Everyone should learn to code with AI! At AI Fund, everyone - not just engineers - can vibe code or use AI assistance to code. This has been great for our creativity and productivity. I hope more teams will empower everyone to build with AI. Please watch the video for details.
This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Daniel Jeffries
@Dan_Jeffries1
"If not for funding for my early work in deep learning from the National Science Foundation (NSF) and Defense Advanced Research Projects Agency (DARPA), which disburse a good deal of U.S. research funding, I would not have discovered lessons about scaling that led me to pitch starting Google Brain to scale up deep learning. I am worried that cuts to funding for basic science will lead the U.S. — and also the world — to miss out on the next set of ideas," writes Andrew Ng.
The government spends a fraction of its budget on basic research and we are about cause a chain reaction of losing down the line because of it.
Each technology is built on the back of the technologies that came before it.
If we don't get clear glass, we don't get eyeglasses, and if we don't get eyeglasses the Dutch eyeglass maker doesn't see his kids using two to magnify things and then we don't get the microscope, and if we don't get the microscope, we don't get germ theory and the marvel of modern medicine.
As James Pethokoukis writes in the Conservative Futurist:
"At the height of the 1960s Space Race, the United States spent nearly 3 percent of its economy on scientific and technological research and development.17 Of that total, about 2 percent was government funding and 1 percent business funding.
"Today, the United States spends...about 0.75 percent by government. (All the various spending bills passed during the first half of the Biden administration—the Bipartisan Infrastructure Act, the CHIPS and Science Act, the Inflation Reduction Act—didn’t change federal R&D investment that much.)
"So especially on the government side, there’s still plenty of room to dramatically ramp up spending without approaching those Apollo-era levels.
"Now there’s no such thing as a free lunch or a money tree. But America’s science and technological innovation system is about as close to having a tasty free lunch beneath a blooming money tree as you’re likely to find.
"Or think of it this way: imagine having a miracle machine where we can put in $1 and get back $5 in the form of higher standards of living, health, and worker productivity, as some economists calculate.
"Of course, the exact product of this machine is hard to predict. Think of the daisy chain of science research and commercial innovation that created Uber.
"It started with Albert Einstein’s special theory of relativity, without which we wouldn’t have the Global Positioning System.19 For GPS to work, it needs to compare extremely accurate time signals from atomic clocks on satellites. But since those satellites are moving at a high velocity compared to Earth-bound Uber app users and experience time differently, they need to be adjusted according to Einstein’s equation. And if no GPS, no Uber with its smartphone-based business model."
We invest in the future or we watch the future pass us by.
Andrew Ng
@AndrewYNg
I am alarmed by the proposed cuts to U.S. funding for basic research, and the impact this would have for U.S. competitiveness in AI and other areas. Funding research that is openly shared benefits the whole world, but the nation it benefits most is the one where the research is done.
If not for funding for my early work in deep learning from the National Science Foundation (NSF) and Defense Advanced Research Projects Agency (DARPA), which disburse a good deal of U.S. research funding, I would not have discovered lessons about scaling that led me to pitch starting Google Brain to scale up deep learning. I am worried that cuts to funding for basic science will lead the U.S. — and also the world — to miss out on the next set of ideas.
In fact, such funding benefits the U.S. more than any other nation. Scientific research brings the greatest benefit to the country where the work happens because (i) the new knowledge diffuses fastest within that country, and (ii) the process of doing research creates new talent for that nation.
Why does most innovation in generative AI still happen in Silicon Valley? Because two teams based in this area — Google Brain, which invented the transformer network, and OpenAI, which scaled it up — did a lot of the early work. Subsequently, team members moved to other nearby businesses, started competitors, or worked with local universities. Further, local social networks rapidly diffused the knowledge through casual coffee meetings, local conferences, and even children’s play dates, where parents of like-aged kids meet and discuss technical ideas. In this way, the knowledge spread faster within Silicon Valley than to other geographies.
In a similar vein, research done in the U.S. diffuses to others in the U.S. much faster than to other geographic areas. This is particularly true when the research is openly shared through papers and/or open source: If researchers have permission to talk about an idea, they can share much more information, such as tips and tricks for how to really make an algorithm work, more quickly. It also lets others figure out faster who can answer their questions. Diffusion of knowledge created in academic environments is especially fast. Academia tends to be completely open, and students and professors, unlike employees of many companies, have full permission to talk about their work.
Thus funding basic research in the U.S. benefits the U.S. most, and also benefits our allies. It is true that openness benefits our adversaries, too. But as a subcommittee of the U.S. House of Representatives committee on science, space, and technology points out, “... open sharing of fundamental research is [not] without risk. Rather, ... openness in research is so important to competitiveness and security that it warrants the risk that adversaries may benefit from scientific openness as well.”
Further, generative AI is evolving so rapidly that staying on the cutting edge is what’s really critical. For example, the fact that many teams can now train a model with GPT-3.5- or even GPT-4-level capability does not seem to be hurting OpenAI much, which is busy growing its business by developing the cutting-edge o4, Codex, GPT-4.1, and so on. Those who invent a technology get to commercialize it first, and in a fast-moving world, the cutting-edge technology is what’s most valuable. Some studies (link in original post, below) also show how knowledge diffuses locally much faster than globally.
China was decisively behind the U.S. in generative AI when ChatGPT was first launched in 2022. However, China’s tech ecosystem is very open internally, and this has helped it to catch up over the past two years:
- There is ample funding for open academic research in China.
- China’s businesses such as DeepSeek and Alibaba have released cutting-edge, open-weights models. This openness at the corporate level accelerates diffusion of knowledge.
- China’s labor laws make non-compete agreements (which stop an employee from jumping ship to a competitor) relatively hard to enforce, and the work culture supports significant idea sharing among employees of different companies; this has made circulation of ideas relatively efficient.
While there’s also much about China that I would not want the U.S. to emulate, the openness of its tech ecosystem has helped it accelerate.
In 1945, Vannevar Bush’s landmark report “Science, The Endless Frontier” laid down key principles for public funding of U.S. research and talent development. Those principles enabled the U.S. to dominate scientific progress for decades. U.S. federal funding for science created numerous breakthroughs that have benefited the U.S. tremendously, and also the world, while training generations of domestic scientists, as well as immigrants who likewise benefit the U.S.
The good news is that this playbook is now well known. I hope many more nations will imitate it and invest heavily in science and talent. And I hope that, having pioneered this very successful model, the U.S. will not pull back from it by enacting drastic cuts to funding scientific research.
[Original post, with links: www.deeplearning.ai/the-batch/issue-303/ ] -
Andrew Ng reposted
Parallæx
@EdgeOfFiRa
The ROI math is staggering. DARPA invested maybe $50M in early deep learning research that created hundreds of billions in market value through Google Brain alone.
Meanwhile we're penny-wise, pound-foolish; debating cuts to programs that historically return 20-30x while China pours money into AI research like water.
Basic research isn't an expense, it's the highest-leverage investment a nation can make.
And great powers seldom fall from external conquest. They fall from internal complacency about the fundamentals that made them great... -
Andrew Ng
@AndrewYNg
I am alarmed by the proposed cuts to U.S. funding for basic research, and the impact this would have for U.S. competitiveness in AI and other areas. Funding research that is openly shared benefits the whole world, but the nation it benefits most is the one where the research is done.
If not for funding for my early work in deep learning from the National Science Foundation (NSF) and Defense Advanced Research Projects Agency (DARPA), which disburse a good deal of U.S. research funding, I would not have discovered lessons about scaling that led me to pitch starting Google Brain to scale up deep learning. I am worried that cuts to funding for basic science will lead the U.S. — and also the world — to miss out on the next set of ideas.
In fact, such funding benefits the U.S. more than any other nation. Scientific research brings the greatest benefit to the country where the work happens because (i) the new knowledge diffuses fastest within that country, and (ii) the process of doing research creates new talent for that nation.
Why does most innovation in generative AI still happen in Silicon Valley? Because two teams based in this area — Google Brain, which invented the transformer network, and OpenAI, which scaled it up — did a lot of the early work. Subsequently, team members moved to other nearby businesses, started competitors, or worked with local universities. Further, local social networks rapidly diffused the knowledge through casual coffee meetings, local conferences, and even children’s play dates, where parents of like-aged kids meet and discuss technical ideas. In this way, the knowledge spread faster within Silicon Valley than to other geographies.
In a similar vein, research done in the U.S. diffuses to others in the U.S. much faster than to other geographic areas. This is particularly true when the research is openly shared through papers and/or open source: If researchers have permission to talk about an idea, they can share much more information, such as tips and tricks for how to really make an algorithm work, more quickly. It also lets others figure out faster who can answer their questions. Diffusion of knowledge created in academic environments is especially fast. Academia tends to be completely open, and students and professors, unlike employees of many companies, have full permission to talk about their work.
Thus funding basic research in the U.S. benefits the U.S. most, and also benefits our allies. It is true that openness benefits our adversaries, too. But as a subcommittee of the U.S. House of Representatives committee on science, space, and technology points out, “... open sharing of fundamental research is [not] without risk. Rather, ... openness in research is so important to competitiveness and security that it warrants the risk that adversaries may benefit from scientific openness as well.”
Further, generative AI is evolving so rapidly that staying on the cutting edge is what’s really critical. For example, the fact that many teams can now train a model with GPT-3.5- or even GPT-4-level capability does not seem to be hurting OpenAI much, which is busy growing its business by developing the cutting-edge o4, Codex, GPT-4.1, and so on. Those who invent a technology get to commercialize it first, and in a fast-moving world, the cutting-edge technology is what’s most valuable. Some studies (link in original post, below) also show how knowledge diffuses locally much faster than globally.
China was decisively behind the U.S. in generative AI when ChatGPT was first launched in 2022. However, China’s tech ecosystem is very open internally, and this has helped it to catch up over the past two years:
- There is ample funding for open academic research in China.
- China’s businesses such as DeepSeek and Alibaba have released cutting-edge, open-weights models. This openness at the corporate level accelerates diffusion of knowledge.
- China’s labor laws make non-compete agreements (which stop an employee from jumping ship to a competitor) relatively hard to enforce, and the work culture supports significant idea sharing among employees of different companies; this has made circulation of ideas relatively efficient.
While there’s also much about China that I would not want the U.S. to emulate, the openness of its tech ecosystem has helped it accelerate.
In 1945, Vannevar Bush’s landmark report “Science, The Endless Frontier” laid down key principles for public funding of U.S. research and talent development. Those principles enabled the U.S. to dominate scientific progress for decades. U.S. federal funding for science created numerous breakthroughs that have benefited the U.S. tremendously, and also the world, while training generations of domestic scientists, as well as immigrants who likewise benefit the U.S.
The good news is that this playbook is now well known. I hope many more nations will imitate it and invest heavily in science and talent. And I hope that, having pioneered this very successful model, the U.S. will not pull back from it by enacting drastic cuts to funding scientific research.
[Original post, with links: www.deeplearning.ai/the-batch/issue-303/ ] -
Andrew Ng
@AndrewYNg
Agentic Document Extraction just got much faster! From previous 135sec median processing time down to 8sec. Extracts not just text but diagrams, charts, and form fields from PDFs to give LLM-ready output. Please see the video for details and some application ideas.
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
In the age of AI, large corporations — not just startups — can move fast. I often speak with large companies’ C-suite and Boards about AI strategy and implementation, and would like to share some ideas that are applicable to big companies. One key is to create an environment where small, scrappy teams don’t need permission to innovate. Let me explain.
Large companies are slower than startups for many reasons. But why are even 3-person, scrappy teams within large companies slower than startups of a similar size? One major reason is that large companies have more to lose, and cannot afford for a small team to build and ship a feature that leaks sensitive information, damages the company brand, hurts revenue, invites regulatory scrutiny, or otherwise damages an important part of the business. To prevent these outcomes, I have seen companies require privacy review, marketing review, financial review, legal review, and so on before a team can ship anything. But if engineers need sign-off from 5 vice presidents before they’re even allowed to launch an MVP (minimum viable product) to run an experiment, how can they ever discover what customers want, iterate quickly, or invent any meaningful new product?
Thanks to AI-assisted coding, the world now has a capability to build software prototypes really fast. But many large companies’ processes – designed to protect against legitimate downside risks – make them unable to take advantage of this capability. In contrast, in small startups with no revenue, no customers, and no brand reputation the downside is limited. In fact, going out of business is a very real possibility anyway, so moving fast makes a superior tradeoff to moving slowly to protect against downside risk. In the worst case, it might invent a new way to go out of business, but in a good case, it might become very valuable.
Fortunately, large companies have a way out of this conundrum. They can create a sandbox environment for teams to experiment in a way that strictly limits the downside risk. Then those teams can go much faster and not have to slow down to get anyone’s permission.
The sandbox environment can be a set of written policies, not necessarily a software implementation of a sandbox. For example, it may permit a team to test the nascent product only on employees of the company and perhaps alpha testers who have signed an NDA, and give no access to sensitive information. It may be allowed to launch product experiments only under newly created brands not tied directly to the company. Perhaps it must operate within a pre-allocated budget for compute.
Within this sandbox, there can be broad scope for experimentation, and — importantly — a team is free to experiment without frequently needing to ask for permission, because the downside they can create is limited. Further, when a prototype shows sufficient promise to bring it to scale, the company can then invest in making sure the software is reliable, secure, treats sensitive information appropriately, is consistent with the company’s brand, and so on.
Under this framework, it is easier to build a company culture that encourages learning, building, and experimentation and celebrates even the inevitable failures that now come with modest cost. Dozens or hundreds of prototypes can be built and quickly discarded as part of the price of finding one or two ideas that turn out to be home runs. This also lets teams move quickly as they churn through those dozens of prototypes needed to get to the valuable ones.
I often speak with large companies about AI strategy and implementation. My quick checklist of things to consider is people, process, and platform. This letter has addressed only part of processes, with an emphasis on moving fast. I’m bullish about what both startups and large companies can do with AI, and I will write about the roles of people and platforms in future letters.
[Original text: www.deeplearning.ai/the-batch/issue-302/ ] -
Andrew Ng
@AndrewYNg
New Course: Reinforcement Fine-Tuning LLMs with GRPO!
Learn to use reinforcement learning to improve your LLM performance in this short course, built in collaboration with @Predibase, and taught by @TravisAddair, its Co-Founder and CTO, and @grg_arnav, its Senior Engineer and Machine Learning Lead.
Reasoning models have been one of the most important developments in LLMs. Reinforcement Fine-Tuning (RFT) uses rewards to encourage LLMs to find solutions to multi-step reasoning tasks such as solving math problems and debugging code - without needing pre-existing training examples like in traditional supervised fine-tuning.
Group Relative Policy Optimization (GRPO) is a reinforcement fine-tuning algorithm gaining rapid adoption. Developed by the DeepSeek team and used to train the R1 reasoning model, GRPO uses reward functions that you can write in Python to assign rewards to model responses. It’s beneficial for tasks with verifiable outcomes and can work well even with fewer than 100 training examples. It can also significantly improve the reasoning ability of smaller LLMs, making applications faster and more cost effective.
In this course, you’ll take a technical deep dive into RFT with GRPO. You’ll learn to build reward functions that you can use in the GRPO training process to guide an LLM toward better performance on multi-step reasoning tasks.
In detail, you’ll:
- Learn when reinforcement fine-tuning is a better fit than supervised fine-tuning, especially for tasks involving multi-step reasoning or limited labeled data.
- Understand how GRPO uses programmable reward functions as a more scalable alternative to the human feedback required for other reinforcement learning algorithms, such as RLHF and DPO.
- Frame the Wordle game as a reinforcement fine-tuning problem and see how an LLM can learn to plan, analyze feedback, and improve its strategy over time.
- Design reward functions that power the reinforcement fine-tuning process.
- Learn techniques for evaluating more subjective tasks, such as rating the quality of a text summary, using an LLM as a judge.
- Understand why reward hacking happens and how to avoid it by adding penalty functions to discourage undesirable behaviors.
- Learn the four key components of the loss calculation in the GRPO algorithm: token probability distribution ratios, advantages, clipping, and KL-divergence.
- Launch reinforcement fine-tuning jobs using Predibase’s hosted training services.
By the end of this course, you’ll be able to build and fine-tune LLMs using reinforcement learning to improve reasoning without relying on large labeled datasets or subjective human feedback.
Please sign up here: www.deeplearning.ai/short-courses/reinforcement-fine-tuni…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI’s ability to make tasks not just cheaper, but also faster, is underrated in its importance in creating business value.
For the task of writing code, AI is a game-changer. It takes so much less effort — and is so much cheaper — to write software with AI assistance than without. But beyond reducing the cost of writing software, AI is shortening the time from idea to working prototype, and the ability to test ideas faster is changing how teams explore and invent. When you can test 20 ideas per month, it dramatically changes what you can do compared to testing 1 idea per month. This is a benefit that comes from AI-enabled speed rather than AI-enabled cost reduction.
That AI-enabled automation can reduce costs is well understood. For example, providing automated customer service is cheaper than operating human-staffed call centers. Many businesses are more willing to invest in growth than just in cost savings; and, when a task becomes cheaper, some businesses will do a lot more of it, thus creating growth. But another recipe for growth is underrated: Making certain tasks much faster (whether or not they also become cheaper) can create significant new value.
I see this pattern across more and more businesses. Consider the following scenarios:
- If a lender can approve loans in minutes using AI, rather than days waiting for a human to review them, this creates more borrowing opportunities (and also lets the lender deploy its capital faster). Even if human-in-the-loop review is needed, using AI to get the most important information to the reviewer might speed things up. The ability to provide loans quickly opens up the market to new customers in need of rapid funds and helps customers who need a quick positive or negative decision to accept the loan or move on.
- If an academic institution gives homework feedback to students in minutes (via sophisticated autograding) rather than days (via human grading), not only is the automation cheaper, the rapid feedback facilitates better learning.
- If an online seller can approve purchases faster, this can lead to more sales. For example, many platforms that accept online ad purchases have an approval process that can take hours or days; if approvals can be done faster, they can earn revenue faster. Further, for customers buying ads, being able to post an ad in minutes lets them test ideas faster and also makes the ad product more valuable.
- If a company’s sales department can prioritize leads and respond to prospective customers in minutes or hours rather than days — closer to when the customers’ buying intent first led them to contact the company — sales representatives might close more deals. Likewise, a business that can respond more quickly to requests for proposals may win more deals.
I’ve written previously about looking at the tasks a company does to explore where AI can help. Many teams already do this with an eye toward making tasks cheaper, either to save costs or to do those tasks many more times. If you’re doing this exercise, consider also whether AI can significantly speed up certain tasks. One place to examine is the sequence of tasks on the path to earning revenue. If some of the steps can be sped up, perhaps this can help revenue growth.
Growth is more interesting to most businesses than cost savings, and if there are loops in your business that, when sped up, would drive growth, AI might be a tool to unlock this growth.
[Original text: www.deeplearning.ai/the-batch/issue-301/ ] -
Andrew Ng
@AndrewYNg
New course: MCP: Build Rich-Context AI Apps with Anthropic. Learn to build AI apps that access tools, data, and prompts using the Model Context Protocol in this short course, created in partnership with Anthropic @AnthropicAI and taught by Elie Schoppik @eschoppik, its Head of Technical Education.
Connecting AI applications to external systems that bring rich context to LLM-based applications has often meant writing custom integrations for each use case. MCP is an open protocol that standardizes how LLMs access tools, data, and prompts from external sources, and simplifies how you provide context to your LLM-based applications. For example, you can provide context via third-party tools that let your LLM make API calls to search the web, access data from local docs, retrieve code from a GitHub repo, and so on.
MCP, developed by Anthropic, is based on a client-server architecture that defines the communication details between an MCP client, hosted inside the AI application, and an MCP server that exposes tools, resources, and prompt templates. The server can be a subprocess launched by the client that runs locally or an independent process running remotely.
In this hands-on course, you'll learn the core architecture behind MCP. You’ll create an MCP-compatible chatbot, build and deploy an MCP server, and connect the chatbot to your MCP server and other open-source servers.
Here’s what you’ll do:
- Understand why MCP makes AI development less fragmented and standardizes connections between AI applications and external data sources
- Learn the core components of the client-server architecture of MCP and the underlying communication mechanism
- Build a chatbot with custom tools for searching academic papers, and transform it into an MCP-compatible application
- Build a local MCP server that exposes tools, resources, and prompt templates using FastMCP, and test it using MCP Inspector
- Create an MCP client inside your chatbot to dynamically connect to your server
- Connect your chatbot to reference servers built by Anthropic’s MCP team, such as filesystem, which implements filesystem operations, and fetch, which extracts contents from the web as markdown
- Configure Claude Desktop to connect to your server and others, and explore how it abstracts away the low-level logic of MCP clients
- Deploy your MCP server remotely and test it with the Inspector or other MCP-compatible applications
- Learn about the roadmap for future MCP development, such as multi-agent architecture, MCP registry API, server discovery, authorization, and authentication
MCP is an exciting and important technology that lets you build rich-context AI applications that connect to a growing ecosystem of MCP servers, with minimal integration work.
Please sign up here! www.deeplearning.ai/short-courses/mcp-build-rich-context-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I’m delighted to announce that AI Fund has closed $190M for our new fund, in an oversubscribed round. I look forward to working with many more builders to create new companies that serve humanity.
AI Fund isn’t a traditional venture capital firm that invests in existing businesses. Instead, we are a venture builder (also called a venture studio): We co-found AI companies, so our team is directly involved in writing code, talking to customers to get feedback, iterating on product designs, preparing market analyses, and so on. We have a lot of fun building multiple AI products at a time, and thus live daily the emerging AI startup best practices.
Many factors go into the success of a startup. But if I had to pick just one, it would be speed. Startups live or die based on their ability to make good decisions and execute fast, which has been a recurring theme of my posts here as well.
If you are building an AI startup, here are some ideas to consider:
- A startup with a small team that pursues one focused, concrete idea can move really fast. Rather than hedging, it is often better to pursue one hypothesis (for example, build one concrete product) but also be willing to switch quickly to a different hypothesis (say, change what features you decide to build) if the data that comes back indicates the original hypothesis is flawed. Concreteness gets you speed!
- A subject matter expert’s gut is remarkably good at making quick decisions. Obviously, there’s a role for data and user studies as well. But if you’re deciding whether to build feature A or B, or to sell first to user persona X or Y, sometimes a domain expert’s gut will point to a quick decision that you can execute and validate or falsify. Trusting a domain expert’s gut gets you speed!
- AI-assisted coding is making prototyping faster than ever before. Yes, AI assistance is speeding up building reliable, enterprise-grade applications and maintaining legacy codebases. But the acceleration it brings to building stand-alone prototypes is far greater. This is because stand-alone prototypes have low requirements for reliability, integration, or even security (if, say, you run them in a sandbox environment). This lets us prototype and test at a ferocious velocity. AI-assisted coding (including vibe coding, where you might barely look at the code) gets you speed!
- Finally, with faster prototyping, the bottleneck shifts to getting feedback from users. A single learning cycle might consist of (i) building a prototype and (ii) getting user feedback to inform the next iteration. Since (i) is now much faster than before, accelerating (ii) is growing in importance. This means teams that are skilled at finding prospective customers and getting their feedback in hours/days rather than weeks can go faster. For example, when building consumer products, I routinely approach strangers (in a respectful way) in public places to ask if they’re willing to give feedback on a prototype I’m working on. (Gathering feedback is more complex for enterprise products, because prospective customers are harder to track down.) Quick user feedback gets you speed!
In addition to speed, a second criterion that I find important for startup success is deep knowledge of the technology. Because AI technology is evolving rapidly, a team with a deep technical understanding of what AI can and cannot do, and when to use what tool, will make better decisions. This creates meaningful differentiation and saves wasting time in blind alleys. A good technical understanding, too, gets you speed!
I’m grateful to AI Fund’s investors, team, and entrepreneur partners for working with us. There is much ahead to build!
[Original text: www.deeplearning.ai/the-batch/issue-300/ ]
Andrew Ng
@AndrewYNg
Additional tips on achieving speed:
- Concreteness: www.deeplearning.ai/the-batch/concrete-ideas-make-strong-…
- Domain expert’s gut: www.deeplearning.ai/the-batch/how-to-brainstorm-ai-startu…
- AI assisted coding: www.deeplearning.ai/the-batch/my-ai-assisted-software-dev…
- Quick user feedback: www.deeplearning.ai/the-batch/how-to-get-user-feedback-to… -
Andrew Ng
@AndrewYNg
Learn to build conversational AI voice agents in "Building AI Voice Agents for Production", created in collaboration with @livekit and @realavatarai, and taught by @dsa (Co-founder & CEO of LiveKit), @shayneparlo (Developer Advocate, LiveKit), and @nedteneva (Head of AI at RealAvatar, an AI Fund portfolio company).
Voice agents combine speech and reasoning capabilities to enable real-time conversations. They're already being used to support customer service, to improve accessibility in healthcare, for entertainment applications, and for talk therapy.
In this course, you’ll learn to build voice agents that listen, reason, and respond naturally. You’ll follow the architecture used to create the "AI Andrew" Avatar, a collaborative project between deeplearning.ai and RealAvatar that responds to users in what sounds like my voice. You’ll build a voice agent from scratch and deploy it to the cloud, enabling support for many simultaneous users.
What you’ll learn:
- Understand the fundamentals of voice agents, including key components like speech-to-text (STT), text-to-speech (TTS), and LLMs, and how latency is introduced at each layer.
- Explore voice agent architectures and the trade-offs between modular pipelines and speech-to-speech APIs.
- Explore how platforms like LiveKit mitigate latency issues with optimized networking infrastructure and low-latency communication protocols.
- Learn how to connect client devices to voice agents using WebRTC—and why it outperforms HTTP and WebSocket for low-latency audio streaming.
- Incorporate voice activity detection (VAD), end-of-turn detection, and context management to detect turns, handle interruptions, and manage conversational flow.
- Understand the trade-offs between latency, quality, and cost in an example in which you build a voice agent and change its voice.
- Equip your agent with metrics to measure latency at each stage of the voice pipeline and learn the key levers you can pull to make your agent faster and more responsive.
The voice agents built in this course also incorporate voice technology from @elevenlabs, a supporting contributor to the project.
By the end of this course, you'll have learned the components of an AI voice agent pipeline, combined them into a system with low-latency communication, and deployed them on cloud infrastructure so it scales to many users.
I’m looking forward to seeing what voice agents you build from this course!
Please sign up here: www.deeplearning.ai/short-courses/building-ai-voice-agent…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
In addition to being a great investor, @WarrenBuffett has also been a great teacher, and I'm grateful to have learned a lot from him. For example, one concept I refer to frequently at @AI_Fund is the Circle of Competence, meaning we figure out what we're good at and what we're not, and act accordingly. His stepping down from Berkshire Hathaway will be the end of an era!
-
Andrew Ng
@AndrewYNg
I hope we can empower everyone to build with AI. Starting from K-12, we should teach every student AI enabled coding, since this will enable them to become more productive and more empowered adults. But there is a huge shortage of computer science (CS) teachers. I recently spoke with high school basketball coach Kyle Creasy, who graduated with a B.A. in Physical Education in 2023. Until two years ago, he had never written a line of Python. Now — with help from AI — he not only writes code, he also teaches CS. I found Kyle’s story inspiring as a model for scaling up CS education in the primary- and secondary-school levels.
Kyle’s success has been with the support of Kira Learning (an AI Fund portfolio company), whose founders Andrea Pasinetti and Jagriti Agrawal have created a compelling vision for CS education. In K-12 classrooms, teachers play a huge social-emotional support role, for example, encouraging students and helping them when they stumble. In addition, they are expected to be subject-matter experts who can deliver the content needed for their subject. Kira Learning uses digital content delivery — educational videos, autograded quizzes, and AI-enabled chatbots to answer students' questions but without giving away homework answers — so the teacher can focus on social-emotional support. While these are still early days, it appears to be working!
A key to making this possible is the hyperpersonalization that is now possible with AI (in contrast to the older idea of the flipped classroom, which had limited adoption). For example, when assigned a problem in an online coding environment, if a student writes this buggy line of Python code
best_$alty_snack = 'potato chips'
Kira Learning’s AI system can spot the problem and directly tell the teacher that $ is an invalid character in a variable name. It can also suggest a specific question for the teacher to ask the student to help get them unstuck, like “Can you identify what characters are allowed in variable names?” Whereas AI can directly deliver personalized advice to students, the fact that it is now helping teachers also deliver personalized support will really help in K-12.
Additionally, agentic workflows can automate a lot of teachers’ repetitive tasks. For example, when designing a curriculum, it’s time-consuming to align the content to educational standards (such as the Common Core in the United States, or the AP CS standard for many CS classes). Having an AI system carry out tasks like these is already proving helpful for teachers.
Since learning to code, Kyle has built many pieces of software. He proudly showed me an analysis he generated in matplotlib of his basketball players’ attempts to shoot three-pointers (shown above), which in turn is affecting the team’s strategy on the court. One lesson is clear: When a basketball coach learns to code, they become a better basketball coach!
I talked about Kyle (and other topics) at the ASU+GSV Summit on education. You can see a video online.
In the future, people who know how to code and build with AI will be much more productive than people who don’t. I’m excited about how AI will lead to new models for K-12 education. By delivering CS education to everyone, I hope that in the future, everyone will be able to build with AI.
[Original text: www.deeplearning.ai/the-batch/issue-299/ ]
Andrew Ng
@AndrewYNg
Video of the talk: www.youtube.com/watch?v=8m4T8ooO5hs
-
Andrew Ng
@AndrewYNg
Even though I’m a much better Python than JavaScript developer, with AI assistance, I’ve been writing a lot of JavaScript code recently. AI-assisted coding, including vibe coding, is making specific programming languages less important, even though learning one is still helpful to make sure you understand the key concepts. This is helping many developers write code in languages we’re not familiar with, which lets us get code working in many more contexts!
My background is in machine learning engineering and back-end development, but AI-assisted coding is making it easy for me to build front-end systems (the part of a website or app that users interact with) using JavaScript (JS) or TypeScript (TS), languages that I am weak in. Generative AI is making syntax less important, so we can all simultaneously be Python, JS, TS, C++, Java, and even Cobol developers. Perhaps one day, instead of being “Python developers" or “C++ developers,” many more of us will just be “developers”!
But understanding the concepts behind different languages is still important. That’s why learning at least one language like Python still offers a great foundation for prompting LLMs to generate code in Python and other languages. If you move from one programming language to another that carries out similar tasks but with different syntax — say, from JS to TS, or C++ to Java, or Rust to Go — once you’ve learned the first set of concepts, you’ll know a lot of the concepts needed to prompt an LLM to code in the second language. (Although TensorFlow and PyTorch are not programming languages, learning the concepts of deep learning behind TensorFlow will also make it much easier to get an LLM to write PyTorch code for you, and vice versa!) In addition, you’ll be able to understand much of the generated code (perhaps with a little LLM assistance).
Different programming languages reflect different views of how to organize computation, and understanding the concepts is still important. For example, someone who does not understand arrays, dictionaries, caches, and memory will be less effective at getting an LLM to write code in most languages.
Similarly, a Python developer who moves toward doing more front-end programming with JS would benefit from learning the concepts behind front-end systems. For example, if you want an LLM to build a front end using the React framework, it will benefit you to understand how React breaks front ends into reusable UI components, and how it updates the DOM data structure that determines what web pages look like. This lets you prompt the LLM much more precisely, and helps you understand how to fix issues if something goes wrong. Similarly, if you want an LLM to help you write code in CUDA or ROCm, it helps to understand how GPUs organize compute and memory.
Just as people who are fluent in multiple human languages can communicate more easily with other people, LLMs are making it easier for developers to build systems in multiple contexts. If you haven’t already done so, I encourage you to try having an LLM write some code in a language you’d like to learn but perhaps haven’t yet gotten around to, and see if it helps you get some new applications to work.
[Original text: www.deeplearning.ai/the-batch/issue-298/ ] -
Andrew Ng
@AndrewYNg
New short course: Building Code Agents with Hugging Face smolagents!
Learn how to build code agents in this course, created in collaboration with @huggingface, and taught by @Thom_Wolf, its co-founder and CSO, and @AymericRoucher, Hugging Face’s Project Lead on Agents.
Tool-calling agents use LLMs to generate multiple function calls sequentially to complete a complex sequence of tasks. They generate one function call, execute it, observe, reason, and decide what to do next. Code agents take a different approach. They consolidate all these calls into a single block of code, letting the LLM lay out an entire action plan at once, which can be executed efficiently to provide more reliable results.
You’ll learn how to code agents using smolagents, a lightweight agentic framework from Hugging Face. Along the way, you’ll learn how to run LLM-generated code safely and develop an evaluation system to optimize your code agent for production.
In detail, you’ll learn:
- How agentic systems have evolved, gaining greater levels of agency over time—and why code agents are a next step.
- How code agents write their actions in code.
- When code agents outperform function-calling agents.
- How to run code agents safely in your system using a constrained Python interpreter and sandboxing using E2B.
- To trace, debug, and assess the code agent to optimize its behaviours for complex requests.
- How to build a research multi-agent system that can find information online and organize it into an interactive report.
By the end of this course, you’ll know how to build and run code agents using smolagents, and deploy them safely with a structured evaluation system in your projects.
Please sign up here! www.deeplearning.ai/short-courses/building-code-agents-wi…This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
AMD
@AMD
👏👏 This week @TIME included our CEO, @LisaSu, as a Titan in their #time100 list! (We may be biased but we think Andrew nailed it.)
“Not every executive can be both grounded and visionary. But Lisa - a former electrical engineer who is charting a transformative course for AMD - continues to make her mark on our world, and all of us are better for it.” - @AndrewYNg
time.com/time100 -
Andrew Ng
@AndrewYNg
I’ve noticed that many GenAI application projects put in automated evaluations (evals) of the system’s output probably later — and rely on humans to judge outputs longer — than they should. This is because building evals is viewed as a massive investment (say, creating 100 or 1,000 examples, and designing and validating metrics) and there’s never a convenient moment to put in that up-front cost. Instead, I encourage teams to think of building evals as an iterative process. It’s okay to start with a quick-and-dirty implementation (say, 5 examples with unoptimized metrics) and then iterate and improve over time. This allows you to gradually shift the burden of evaluations away from humans and toward automated evals.
I wrote previously in The Batch about the importance and difficulty of creating evals. Say you’re building a customer-service chatbot that responds to users in free text. There’s no single right answer, so many teams end up having humans pore over dozens of example outputs with every update to judge if it improved the system. While techniques like LLM-as-judge are helpful, the details of getting this to work well (such as what prompt to use, what context to give the judge, and so on) are finicky to get right. All this contributes to the impression that building evals requires a large up-front investment, and thus on any given day, a team can make more progress by relying on human judges than figuring out how to build automated evals.
I encourage you to approach building evals differently. It’s okay to build quick evals that are only partial, incomplete, and noisy measures of the system’s performance, and to iteratively improve them. They can be a complement to, rather than replacement for, manual evaluations. Over time, you can gradually tune the evaluation methodology to close the gap between the evals’ output and human judgments. For example:
- It’s okay to start with very few examples in the eval set, say 5, and gradually add to them over time — or subtract them if you find that some examples are too easy or too hard, and not useful for distinguishing between the performance of different versions of your system.
- It’s okay to start with evals that measure only a subset of the dimensions of performance you care about, or measure narrow cues that you believe are correlated with, but don’t fully capture, system performance. For example if, at a certain moment in the conversation, your customer-support agent is supposed to (i) call an API to issue a refund and (ii) generate an appropriate message to the user, you might start off measuring only whether or not it calls the API correctly and not worry about the message. Or if, at a certain moment, your chatbot should recommend a specific product, a basic eval could measure whether or not the chatbot mentions that product without worrying about what it says about it.
So long as the output of the evals correlates with overall performance, it’s fine to measure only a subset of things you care about when starting.
The development process thus comprises two iterative loops, which you might execute in parallel:
- Iterating on the system to make it perform better, as measured by a combination of automated evals and human judgment;
- Iterating on the evals to make them correspond more closely to human judgment.
As with many things in AI, we often don’t get it right the first time. So t’s better to build an initial end-to-end system quickly and then iterate to improve it. We’re used to taking this approach to building AI systems. We can build evals the same way.
To me, a successful eval meets the following criteria. Say, we currently have system A, and we might tweak it to get a system B:
- If A works significantly better than B according to a skilled human judge, the eval should give A a significantly higher score than B.
- If A and B have similar performance, their eval scores should be similar.
Whenever a pair of systems A and B contradicts these criteria, that is a sign the eval is in “error” and we should tweak it to make it rank A and B correctly. This is a similar philosophy to error analysis in building machine learning algorithms, only instead of focusing on errors of the machine learning algorithm's output — such as when it outputs an incorrect label — we focus on “errors” of the evals — such as when they incorrectly rank two systems A and B, so the evals aren’t helpful in choosing between them.
Relying purely on human judgment is a great way to get started on a project. But for many teams, building evals as a quick prototype and iterating to something more mature lets you put in evals earlier and accelerate your progress.
[Original text: www.deeplearning.ai/the-batch/issue-297/ ] -
Andrew Ng
@AndrewYNg
New Short Course: Building AI Browser Agents!
Learn how to build AI agents that interact and take actions on websites in this course, created in partnership with @agi_inc and taught by @divgarg and @namangarg0, Co-founders of AGI Inc.
AI browser agents can log into websites, fill out forms, click through web pages, or even place orders online for you. They use both visual information, like screenshots, and structural data, like the HTML or Document Object Model (DOM) of a web page, to reason and take action.
With the complexity of webpages and multiple possible actions at each step, it can be challenging for an AI browser agent to complete an assigned task. Because these agents run long action sequences, a single error—like clicking the wrong button or misreading a field—can lead to unexpected outcomes or errors that compound over time.
In this course, you'll understand how autonomous web agents work, their current limitations, and how AgentQ enables them to improve through self-correction.
In detail, you'll:
- Learn what web agents are, how they automate tasks online, their architecture, key components, limitations, and an overview of their decision-making strategies.
- Build a web agent that can scrape deeplearning.ai's website and return course recommendations in a structured output format.
- Build an autonomous web agent that can execute multiple tasks, such as finding and summarizing webpages, filling out a form, and signing up for a newsletter.
- Explore AgentQ, a framework that enables agents to self-correct by combining Monte Carlo Tree Search (MCTS), a self-critique mechanism for continuous improvement, and Direct Preference Optimization (DPO).
- Deep dive into MCTS, learn how it finds an effective path, illustrated by an example of Gridworld animation, and use AgentQ to complete web tasks.
- Understand AI agents' current state and future directions—including key factors shaping their evolution, such as hardware, algorithm innovation, and data availability.
By the end of this course, you will have hands-on experience building browser agents and a deeper understanding of how to make them more robust and reliable.
Please sign up here: www.deeplearning.ai/short-courses/building-ai-browser-age…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I am so sorry that the U.S. is letting down our friends and allies. Broad tariffs, implemented not just against adversaries but also steadfast allies, will damage the livelihoods of billions of people, create inflation, make the world more fragmented, and leave the U.S. and the world poorer. AI isn’t the solution to everything, but even amidst this challenging environment, I hope our community can hold together, keep building friendships across borders, keep sharing ideas, and keep supporting each other.
Much has been written about why high, widespread taxes on imports are harmful. In this letter, I’d like to focus on its possible effects on AI. One silver lining of the new tariffs is that they focus on physical imports, rather than digital goods and services, including intellectual property (IP) such as AI research inventions and software. IP is difficult to tax, because each piece of IP is unique and thus hard to value, and it moves across borders with little friction via the internet. Many international AI teams collaborate across borders and timezones, and software, including specifically open source software, is an important mechanism for sharing ideas. I hope that this free flow of ideas remains unhampered, even if the flow of physical goods is.
However, AI relies on hardware, and tariffs will slow down AI progress by restricting access to it. Even though a last-minute exception was made for semiconductors, taxing imports of solar panels, wind turbines, and other power-generation and -distribution equipment will diminish the ability to provide power to U.S. data centers. Taxing imports of servers, cooling hardware, networking hardware, and the like will also make it more expensive to build data centers. And taxing consumer electronics, like laptops and phones, will make it harder for citizens to learn and use AI.
With regard to data-center buildouts, another silver lining is that, with the rise of generative AI, data gravity has decreased because compute processing costs are much greater than transmission costs, meaning it’s more feasible to place data centers anywhere in the world rather than only in close proximity to end-users. Even though many places do not have enough trained technicians to build and operate data centers, I expect tariffs will encourage data centers to be built around the world, creating more job opportunities globally.
Finally, tariffs will create increased pressure for domestic manufacturing, which might create very mild tailwinds for robotics and industrial automation. As U.S. Vice President J.D. Vance pointed out in 2017, the U.S. should focus on automation (and education) rather than on tariffs. But the U.S. does not have the personnel — or know-how, or supply chain — to manufacture many of the goods that it currently counts on allies to make. Robotics can be helpful for addressing a small part of this large set of challenges. Generative AI’s rate of progress in robotics is also significantly slower than in processing text, visual data, audio, and reasoning. So while the tariffs could create tailwinds for AI-enabled robotics, I expect this effect to be small.
My 4-year-old son had been complaining for a couple of weeks that his shoes were a tight fit — he was proud that he’s growing! So last Sunday, we went shoe shopping. His new shoes cost $25, and while checking out, I paused and reflected on how lucky I am to be able to afford them. But I also thought about the many families living paycheck-to-paycheck, and for whom tariffs leading to shoes at $40 a pair would mean they let their kids wear ill-fitting shoes longer. I also thought about people I’ve met in clothing manufacturing plants in Asia and Latin America, for whom reduced demand would mean less work and less money to take home to their own kids.
I don’t know what will happen next with the U.S. tariffs, and plenty of international trade will happen with or without U.S. involvement. I hope we can return to a world of vibrant global trade with strong, rules-based, U.S. participation. Until then, let’s all of us in AI keep nurturing our international friendships, keep up the digital flow of ideas — including specifically open source software — and keep supporting each other. Let’s all do what we can to keep the world as connected as we are able.
[I had written this letter before the 90 day pause on the tariffs, but am sharing this here since many of the points are still relevant depends on what happens next.]
Original text: www.deeplearning.ai/the-batch/issue-296/ -
Andrew Ng
@AndrewYNg
Contrary to standard prompting advice that you should give LLMs the context they need to succeed, I find it’s sometimes faster to be lazy and dash off a quick, imprecise prompt and see what happens. The key to whether this is a good idea is whether you can quickly assess the output quality, so you can decide whether to provide more context. In this post, I’d like to share when and how I use “lazy prompting.”
When debugging code, many developers copy-paste error messages — sometimes pages of them — into an LLM without further instructions. Most LLMs are smart enough to figure out that you want them to help understand and propose fixes, so you don’t need to explicitly tell them. With brief instructions like “Edit this: …” or “sample dotenv code” (to remind you how to write code to use Python's dotenv package), an LLM will often generate a good response. Further, if the response is flawed, hopefully you can spot any problems and refine the prompt, for example to steer how the LLM edits your text.
At the other end of the spectrum, sometimes I spend 30 minutes carefully writing a 2-page prompt to get an AI system to help me solve a problem (for example to write many pages of code) that otherwise would have taken me much longer.
I don’t try a lazy prompt if (i) I feel confident there’s no chance the LLM will provide a good solution without additional context. For example, given a partial program spec, does even a skilled human developer have a chance of understanding what you want? If I absolutely want to use a particular piece of pdf-to-text conversion software (like my team LandingAI’s Agentic Doc Extraction!), I should say so in the prompt, since otherwise it’s very hard for the LLM to guess my preference. I also wouldn’t use a lazy prompt if (ii) a buggy implementation would take a long time to detect. For example, if the only way for me to figure out if the output is incorrect is to laboriously run the code to check its functionality, it would be better to spend the time up-front to give context that would increase the odds of the LLM generating what I want.
By the way, lazy prompting is an advanced technique. On average, I see more people giving too little context to LLMs than too much. Laziness is a good technique only when you’ve learned how to provide enough context, and then deliberately step back to see how little context you can get away with and still have it work. Also, lazy prompting applies only when you can iterate quickly using an LLM’s web or app interface. It doesn’t apply to prompts written in code for the purpose of repeatedly calling an API, since presumably you won’t be examining every output to clarify and iterate if the output is poor.
Thank you to Rohit Prsad, who has been collaborating with me on the open-source package aisuite, for suggesting the term lazy prompting. There is an analogy to lazy evaluation in computer science, where you call a function at the latest possible moment and only when a specific result is needed. In lazy prompting, we add details to the prompt only when they are needed.
Original text: www.deeplearning.ai/the-batch/issue-295/ -
Andrew Ng
@AndrewYNg
New Short Course: Getting Structured LLM Output!
Learn how to get structured outputs from your LLM applications in this course, built in partnership with @dottxtai, and taught by @willkurt, a Founding Engineer, and @just_cameron , Developer Relations Engineer.
It's challenging for software to automatically parse through an LLM's freeform text outputs. Structured outputs—like JSON—solve this by converting natural language into consistent, clear, data that a machine can read and process. This course teaches you how to generate structured outputs while building several use cases, including a social media analysis agent.
You’ll learn about structured outputs and efficient ways to generate outputs in your defined schema or format. You’ll begin by using structured output APIs, then use re-prompting libraries like “instructor” to generate structured output. Finally, you’ll learn how constrained decoding works; this is a very clever technique in which constraints are applied on each subsequent token generated, blocking any tokens that don’t fit your defined schema.
In detail, you’ll:
- Learn why structured outputs are important, how they allow for scalable software development, and the different approaches to generate them, including vendor-provided APIs, re-prompting libraries, and structured generation.
- Build a simple social media agent using OpenAI’s structured output API, learn how to define a model's desired structured output using Pydantic, and perform basic programming with your outputs, such as importing structured data into a data frame using pandas.
- Learn how to use the open-source library "instructor," which checks the structured output of the model and re-prompts the model until it validates the desired output, and explore the limitations of this approach.
- Understand how structured generation by the “outlines” library works by modifying LLM logits, on a per-generated-token basis based on the desired format, to give a particular output structure.
- Learn how regular expressions, which outlines works with, are represented as finite-state machines, and how they can be used to develop a range of structured outputs beyond JSON.
By the end of this course, you’ll have broadened your knowledge of the approaches you can use to get structured outputs from your LLM applications.
Please sign up here: www.deeplearning.ai/short-courses/getting-structured-llm-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Major program launch: Data Analytics Professional Certificate! This large, five-course sequence takes you all the way to being job-ready as a data analyst, and shows how to use Generative AI as a thought partner to enhance your work in this role.
Offered by deeplearning.ai on Coursera, this is taught by Sean Barnes, Ph.D., a Data Science & Engineering Leader at Netflix.
Analyzing data remains one of the most important skills in where the world is going with AI. This comprehensive certificate takes you all the way to being job-ready.
Each course comes with practical projects demonstrated in real-world contexts, such as analyzing sales data for a Korean bakery, video game sales trends across different regions, or identifying factors impacting customer retention for a communications company. You'll also work on estimating fire distribution for forest fire prevention, analyzing how a diamond's properties affect its market value, and developing predictive models for retail sales analysis, carbon emissions, and coral reef conservation.
Here's some of what you'll learn:
- How to define data and categorize it into its many types such as discrete & continuous numerical, structured & unstructured, time series, categorical, and know what insights can be derived from the different types of data categories.
- How to differentiate between data-related job roles and their responsibilities, and how data flows through an organization from the moment of capture to decision-making.
- How to perform data processing functions and apply conditional formatting in spreadsheets to extract business value from your data using statistical calculations and best practices for visualizing and interpreting data.
- How to use LLMs for stakeholder analysis, data exploration, and data visualization.
- Best practices for using LLMs for as a thought partner to data analysis work
By the end of this professional certificate program, you will have learned core statistical concepts, analysis techniques, and visualization methodologies that will serve as the foundation for working as a data analyst.
The world needs more data analysts, especially ones who know how to use modern generative AI. With data science roles projected to grow 36% by 2033, the skills taught in this program create new professional opportunities in data.
Sign up here! www.coursera.org/professional-certificates/data-analyticsThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course: Vibe Coding 101 with Replit! Learn to build and host applications with an AI agent in this course, built in partnership with @Replit and taught by its President @pirroh and Head of Developer Relations @mattyp.
Coding agents are changing how we write code. "Vibe coding" refers to a growing practice where you might barely look at the generated code, and instead focus on the architecture and features of your application. However, contrary to popular belief, effectively coding this way isn't done by just prompting, accepting all recommendations, and hoping for the best. It requires structuring your work, refining your prompts, and having a systematic process that lead to a more efficient and effective workflow.
I code frequently using LLMs, and asking an LLM to do everything in one shot usually does not work. I'll typically take a problem, partition it into manageable modules, spend time creating prompts to specify each module, and use the model to produce the code one module at a time, and test/debug each module before moving on. A process like this is making me and many other developers faster and more efficient.
In this video-only course, you’ll learn how to use Replit’s cloud environment--with an integrated code editor, package manager, and deployment tools--to build and deploy web applications. Along the way, you’ll learn strategies for working effectively with agents and improve your development skills.
In detail, you’ll:
- Understand principles of agentic code development such as being precise, giving agents one task at a time, making prompts specific, keeping projects tidy, starting with fresh sessions for each new feature, and how to approach debugging.
- Learn how to get started with Replit, and key skills for vibe coding: Thinking, using frameworks, checkpoints, debugging, and providing context.
- Create a product requirement document (PRD) and wireframe for your agent to build a prototype of a website performance analyzer.
- See how to use an agent to make your prototype more visually appealing, and deploy it application others to access .
- Learn to build a head-to-head national park ranking app, from a sample dataset, with voting capabilities and persistent data storage, and refine further ask the assistant to recap and explain what it built to find room for improvement and reinforce your learning.
By the end of this course, you’ll have a solid foundation in building with coding agents, and a process you can use to keep vibe coding effectively.
Please sign up here: www.deeplearning.ai/short-courses/vibe-coding-101-with-re…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Last Friday on Pi Day, we held AI Dev 25, a new conference for AI Developers. Tickets had (unfortunately) sold out shortly after we announced their availability, but I came away energized by the day of coding and technical discussions with fellow AI Builders! Let me share here my observations from the event.
I'd decided to start AI Dev because while there're great academic AI conferences that disseminate research work (such as NeurIPS, ICML and ICLR) and also great meetings held by individual companies, often focused on each company's product offerings, there were few vendor-neutral conferences for AI developers. With the wide range of AI tools now available, there is a rich set of opportunities for developers to build new things (and to share ideas on how to build things!), but also a need for a neutral forum that helps developers do so.
Based on an informal poll, about half the attendees had traveled to San Francisco from outside the Bay Area for this meeting, including many who had come from overseas. I was thrilled by the enthusiasm to be part of this AI Builder community. To everyone who came, thank you!
Other aspects of the event that struck me:
- First, agentic AI continues to be a strong theme. The topic attendees most wanted to hear about (based on free text responses to our in-person survey at the start of the event) was agents!
- Google's Paige Bailey talked about embedding AI in everything and using a wide range of models to do so. I also particularly enjoyed her demos of Astra and Deep Research agents.
- Meta's Amit Sangani talked compellingly as usual about open models. Specifically, he described developers fine-tuning smaller models on specific data, resulting in superior performance than with large general purpose models. While there're still many companies using fine-tuning that should really just be prompting, I'm also seeing continued growth of fine-tuning in applications that are reaching scale and that are becoming valuable.
- Many speakers also spoke about the importance of being pragmatic about what problems we are solving, as opposed to buying into the AGI hype. For example, Nebius' Roman Chernin put it simply: Focusing on solving real problems is important!
- Lastly, I was excited to hear continued enthusiasm for the Voice Stack. Justin Uberti gave a talk about OpenAI’s realtime audio API to a packed room, with many people pulling out laptops to try things out themselves in code!
deeplearning.ai has a strong “Learner First” mentality; our foremost goal is always to help learners. I was thrilled that a few attendees told me they enjoyed how technical the sessions were, and said they learned many things that they're sure they will use. (In fact, I, too, came away with a few ideas from the sessions!) I was also struck that, both during the talks and at the technical demo booths, the rooms were packed with attendees who were highly engaged throughout the whole day. I'm glad that we were able to have a meeting filled with technical and engineering discussions.
I'm delighted that AI Dev 25 went off so well, and am grateful to all the attendees, volunteers, speakers, sponsors, partners, and team members that made the event possible. I regretted only that the physical size of the event space prevented us from admitting more attendees this time. There is something magical about bringing people together physically to share ideas, make friends, and to learn from and help each other. I hope we'll be able to bring even more people together in the future.
[Original text: www.deeplearning.ai/the-batch/issue-293/ ]This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course: Long-Term Agentic Memory with LangGraph. Learn to build an agent with long-term memory in this course developed in collaboration with @langchain taught by its Co-Founder and CEO, @hwchase17!
Personal assistance and productivity tasks have become important use cases for agents. An important feature of an AI assistant, such as a coding or calendar assistant, is its ability to keep improving over time from its experience. Agent memory is the key capability that enables this.
To add memory to an agent, you must first figure out what to store and what to retrieve when it is time to use the information. Additionally, you’ll have to decide when to update the stored information. For example, you might update in each iteration loop of the agent or perform updates in the background, with a helper agent.
In this course, you will learn a mental framework to build agents with long-term memory. You'll create a useful email assistant that can respond, ignore, and notify using writing, scheduling, and memory-management tools. You’ll develop your agent's memory by adding facts to its memory store, provide examples to learn the user's preferences, and optimize system prompts to evolve instructions based on previous responses.
In detail, you’ll:
- Learn how the three types of memory--semantic, episodic, and procedural–and the two update mechanisms–via hot path and in the background–apply to your agents.
- Build an email agent with writing, scheduling, and availability tools, along with a router that triages incoming email and handles it accordingly by ignoring, responding, or notifying the user.
- Add tools to your email agent that allow it to operate on semantic memory by learning facts about the user, storing them in a long-term memory store, and searching over them in future interactions.
- Incorporate episodic memory, in the form of few-shot examples, in the triage step of your agents to help them learn and update user preferences.
- Add procedural memory as system prompts, optimized with feedback to improve the instructions the agent follows.
Learn how to approach memory in agents, and start building agents with long-term memory with LangGraph!
Please sign up here: www.deeplearning.ai/short-courses/long-term-agentic-memor…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
OpenAI’s Justin Uberti at AI Dev 25 showing how to build a voice agent using the Realtime API. Building on the voice stack is easier than most realize - worth trying out if you have a voice idea!
-
Andrew Ng
@AndrewYNg
Good tip from Replit’s @mattyp at AI Dev 25 on debugging while vibe coding: Large part of it is looking at outputs to figure out what context you have that LLM does not, so that you can give it that context and help it get unstuck. Sometimes pasting in the error messages is enough, but also sometimes not.
-
Andrew Ng
@AndrewYNg
It’s starting - just kicked off AI Dev 25, the AI developer conference, in San Francisco! Happy Pi day!
Andrew Ng
@AndrewYNg
From audience poll the topic AI developers are most excited about is Agents!
Andrew Ng
@AndrewYNg
Great talk by Google’s Bill Jia on their GenAI work, including Astra and Deep Research agents (both of which I think are very cool).
Andrew Ng
@AndrewYNg
Panel with Replit’s Michele Catasta, Stanford’s @percyliang , Nebius’ Roman Chernin and Hugging Face’s @Thom_Wolf, moderated by @lmoroney, on application building. Lots of tips on infra, open source, agentic workflows, benchmarking and code gen. Particular interest in how to take stochastic LLMs that hallucinate and nonetheless build reliable agents.
Andrew Ng
@AndrewYNg
Meta’s Chaya Nayak talking about the open Llama models and Llama Stack, and best practices for using them. Great tips and I saw lots of people pulling out phones to take pictures of her slides!
-
Andrew Ng
@AndrewYNg
Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the programming occupation will become extinct [...] than that it will become all-powerful. More and more, computers will program themselves.” Statements discouraging people from learning to code are harmful!
In the 1960s, when programming moved from punchcards (where a programmer had to laboriously make holes in physical cards to write code character by character) to keyboards with terminals, programming became easier. And that made it a better time than before to begin programming. Yet it was in this era that Nobel laureate Herb Simon wrote the words quoted in the first paragraph. Today’s arguments not to learn to code continue to echo his comment.
As coding becomes easier, more people should code, not fewer!
Over the past few decades, as programming has moved from assembly language to higher-level languages like C, from desktop to cloud, from raw text editors to IDEs to AI assisted coding where sometimes one barely even looks at the generated code (which some coders recently started to call vibe coding), it is getting easier with each step.
I wrote previously that I see tech-savvy people coordinating AI tools to move toward being 10x professionals — individuals who have 10 times the impact of the average person in their field. I am increasingly convinced that the best way for many people to accomplish this is not to be just consumers of AI applications, but to learn enough coding to use AI-assisted coding tools effectively.
One question I’m asked most often is what someone should do who is worried about job displacement by AI. My answer is: Learn about AI and take control of it, because one of the most important skills in the future will be the ability to tell a computer exactly what you want, so it can do that for you. Coding (or getting AI to code for you) is a great way to do that.
When I was working on the course Generative AI for Everyone and needed to generate AI artwork for the background images, I worked with a collaborator who had studied art history and knew the language of art. He prompted Midjourney with terminology based on the historical style, palette, artist inspiration and so on — using the language of art — to get the result he wanted. I didn’t know this language, and my paltry attempts at prompting could not deliver as effective a result.
Similarly, scientists, analysts, marketers, recruiters, and people of a wide range of professions who understand the language of software through their knowledge of coding can tell an LLM or an AI-enabled IDE what they want much more precisely, and get much better results. As these tools are continuing to make coding easier, this is the best time yet to learn to code, to learn the language of software, and learn to make computers do exactly what you want them to do.
[Original text: www.deeplearning.ai/the-batch/issue-292/ ] -
Andrew Ng
@AndrewYNg
Continuing from last week’s post on the rise of the Voice Stack, there’s an area that today’s voice-based systems often struggle with: Voice Activity Detection (VAD) and the turn-taking paradigm of communication.
When communicating with a text-based chatbot, the turns are clear: You write something, then the bot does, then you do, and so on. The success of text-based chatbots with clear turn-taking has influenced the design of voice-based bots, most of which also use the turn-taking paradigm.
A key part of building such a system is a VAD component to detect when the user is talking. This allows our software to take the parts of the audio stream in which the user is saying something and pass that to the model for the user’s turn. It also supports interruption in a limited way, whereby if a user insistently interrupts the AI system while it is talking, eventually the VAD system will realize the user is talking, shut off the AI’s output, and let the user take a turn. This works reasonably well in quiet environments.
However, VAD systems today struggle with noisy environments, particularly when the background noise is from other human speech. For example, if you are in a noisy cafe speaking with a voice chatbot, VAD — which is usually trained to detect human speech — tends to be inaccurate at figuring out when you, or someone else, is talking. (In comparison, it works much better if you are in a noisy vehicle, since the background noise is more clearly not human speech.) It might think you are interrupting when it was merely someone in the background speaking, or fail to recognize that you’ve stopped talking. This is why today’s speech applications often struggle in noisy environments.
Intriguingly, last year, Kyutai Labs published Moshi, a model that had many technical innovations. An important one was enabling persistent bi-direction audio streams from the user to Moshi and from Moshi to the user.
If you and I were speaking in person or on the phone, we would constantly be streaming audio to each other (through the air or the phone system), and we’d use social cues to know when to listen and how to politely interrupt if one of us felt the need. Thus, the streams would not need to explicitly model turn-taking. Moshi works like this. It’s listening all the time, and it’s up to the model to decide when to stay silent and when to talk. This means an explicit VAD step is no longer necessary. (Moshi also included other innovations, such as an “inner monologue” that simultaneously generates text alongside the audio to improve the quality of responses as well as audio encoding.)
Just as the architecture of text-only transformers has gone through many evolutions (such as encoder-decoder models, decoder-only models, and reasoning models that generate a lot of “reasoning tokens” before the final output), voice models are going through a lot of architecture explorations. Given the importance of foundation models with voice-in and voice-out capabilities, many large companies right now are investing in developing better voice models. I’m confident we’ll see many more good voice models released this year.
It feels like the space of potential innovation for voice remains large. Hard technical problems, like the one of latency that I described last week and VAD errors, remain to be solved. As solutions get better, voice-to-voice will continue to be a promising category to build applications in.
[Original text: www.deeplearning.ai/the-batch/issue-291/ ] -
Andrew Ng
@AndrewYNg
New short course: Event-Driven Agentic Document Workflows. Event-driven workflows are a key design pattern in which many pieces of work (LLM calls, tool use, etc.) can be carried out asynchronously and in parallel, and completion of specific steps generate events that trigger other work to begin. In this course, created in partnership with @llama_index and @seldo, VP of Developer Relations, you'll learn to apply this technique to document workflows.
Filling out complex forms can be tedious, time-consuming, and error-prone. Agentic workflows can automate this. This course teaches you how to build an agentic document workflow using an event-driven architecture.
You'll design an event-driven agentic workflow that fills out a PDF form based on information from a source document. The agent will use RAG to retrieve relevant data from the source document, parse the form to identify the required fields, convert the blank spaces into questions, and send those questions to the RAG system to fill the form. You'll collaborate with the agent using a human-in-the-loop feedback approach through text and with your voice.
In detail, you’ll:
- Understand the basic concepts of event-driven workflows.
- Build a series of LlamaIndex’s workflows that increase in complexity from branching and looping logic to concurrent executions.
- Set up the agent’s RAG capability by parsing the source document, loading the extracted information into a vector store, and building a query engine on top of the store.
- Implement workflow steps that enable the agent to parse the form to be filled, turn the parsed information into simple questions, and use the questions to query the RAG pipeline.
- Incorporate human-in-the-loop into the workflow and ask the agent to re-answer when necessary to produce more accurate form responses.
- Add multimodal capability to the agent, allowing it to process spoken feedback.
By the end of this course, you will have built an event-driven agentic workflow that fills out a document and responds to human feedback to complete the form more accurately.
Please sign up here: www.deeplearning.ai/short-courses/event-driven-agentic-do…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Announcing: Agentic Document Extraction!
PDF files represent information visually - via layout, charts, graphs, etc. - and are more than just text. Unlike traditional OCR and most PDF-to-text approaches, which focus on extracting the text, an agentic approach lets us break a document down into components and reason about them, resulting in more accurate extraction of the underlying meaning for RAG and other applications. Watch the video for details.This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
The Voice Stack is improving rapidly. Systems that interact with users via speaking and listening will drive many new applications. Over the past year, I’ve been working closely with deeplearning.ai, AI Fund, and several collaborators on voice-based applications, and I will share best practices I’ve learned in this and future posts.
Foundation models that are trained to directly input, and often also directly generate, audio have contributed to this growth, but they are only part of the story. OpenAI’s RealTime API makes it easy for developers to write prompts to develop systems that deliver voice-in, voice-out experiences. This is great for building quick-and-dirty prototypes, and it also works well for low-stakes conversations where making an occasional mistake is okay. I encourage you to try it!
However, compared to text-based generation, it is still hard to control the output of voice-in voice-out models. In contrast to directly generating audio, when we use an LLM to generate text, we have many tools for building guardrails, and we can double-check the output before showing it to users. We can also use sophisticated agentic reasoning workflows to compute high-quality outputs. Before a customer-service agent shows a user the message, “Sure, I’m happy to issue a refund,” we can make sure that (i) issuing the refund is consistent with our business policy and (ii) we will call the API to issue the refund (and not just promise a refund without issuing it).
In contrast, the tools to prevent a voice-in, voice-out model from making such mistakes are much less mature.
In my experience, the reasoning capability of voice models also seems inferior to text-based models, and they give less sophisticated answers. (Perhaps this is because voice responses have to be more brief, leaving less room for chain-of-thought reasoning to get to a more thoughtful answer.)
When building applications where I need a high degree of control over the output, I use agentic workflows to reason at length about the user’s input. In voice applications, this means I end up using a pipeline that includes speech-to-text (STT, also known as ASR, or automatic speech recognition) to transcribe the user’s words, then processes the text using one or more LLM calls, and finally returns an audio response to the user via TTS (text-to-speech). This STT → LLM/Agentic workflow → TTS pipeline, where the reasoning is done in text, allows for more accurate responses.
However, this process introduces latency, and users of voice applications are very sensitive to latency. When deeplearning.ai worked with RealAvatar (an AI Fund portfolio company led by Jeff Daniel) to build an avatar of me, we found that getting TTS to generate a voice that sounded like me was not very hard, but getting it to respond to questions using words similar to those I would choose was. Even after a year of tuning our system — starting with iterating on multiple, long, mega-prompts and eventually developing complex agentic workflows — it remains a work in progress. You can play with it at deeplearning.ai/avatar
Initially, this agentic workflow incurred 5-9 seconds of latency, and having users wait that long for responses led to a bad experience. To address this, we came up with the following latency reduction technique. The system quickly generates a pre-response (short for preliminary response) that can be uttered quickly, which buys time for an agentic workflow to generate a more thoughtful, full response. (We’re grateful to LiveKit’s CEO Russ d’Sa and team for helping us get this working.) This is similar to how, if you were to ask me a complicated question, I might say “Hmm, let me think about that” or “Sure, I can help with that” — that’s the pre-response — while thinking about what my full response might be.
I think generating a pre-response followed by a full response, to quickly acknowledge the user’s query and also reduce the perceived latency, will be an important technique, and I hope many teams will find this useful. Our goal was to approach human face-to-face conversational latency, which is around 0.3-1 seconds. RealAvatar and deeplearning.ai, through our efforts on the pre-response and other optimizations, have reduced the system’s latency to around 0.5-1 seconds.
Months ago, sitting in a coffee shop, I was able to buy a phone number on Twilio and hook it up to an STT → LLM → TTS pipeline in just hours. This enabled me to talk to my own LLM using custom prompts. Prototyping voice applications is much easier than most people realize!
Building reliable, scaled production applications takes longer, of course, but if you have a voice application in mind, I hope you’ll start building prototypes and see how far you can get! I’ll keep building voice applications and sharing best practices and voice-related technology trends in future posts.
[Original letter: www.deeplearning.ai/the-batch/issue-290/ ] -
Andrew Ng
@AndrewYNg
Transformers have dominated LLM text generation, and generate tokens sequentially. This is a cool attempt to explore diffusion models as an alternative, by generating the entire text at the same time using a coarse-to-fine process. Congrats @StefanoErmon & team!
Inception
@_inception_ai
We are excited to introduce Mercury, the first commercial-grade diffusion large language model (dLLM)! dLLMs push the frontier of intelligence and speed with parallel, coarse-to-fine text generation.
-
Andrew Ng
@AndrewYNg
New course to bring you up to state-of-the-art at using AI to help you code: Build Apps with Windsurf's AI Coding Agents, built in partnership with WIndsurf (@codeiumdev) and taught by @_anshulr!
AI-assisted IDEs (Integrated Development Environments) make developers’ workflows faster, more efficient, and much more fun. Agentic tools like Windsurf are more than just code autocomplete—they are collaborative coding agents that help you break down complex applications, iterate efficiently, and generate code that spans multiple files.
Although a lot of coding assistants share the same underlying large language models for planning and reasoning, a major point of distinction is how they handle tools, keep track of context, and stay aligned with your intent as a developer.
For instance, if you make modifications to a class definition in your code and make the same modifications to other classes in the same directory, you might tell the AI agent "Do the same thing in similar places in this directory." Here, tracking your intent means understanding that “the same thing" refers to that recent edit you just made, which must be followed by appropriate search and tool-calling to implement the changes.
In this course, you'll learn the inner workings of coding agents, their strengths and limitations, and how to use Windsurf to quickly build several applications.
In detail, you'll:
- Build a mental model of how agents work by combining human-action tracking, tool integration, and context awareness to carry out an agentic coding workflow.
- Learn the challenges of code search and discovery and how a multi-step retrieval approach helps coding agents address them.
- Use Windsurf to analyze and understand a large, old codebase and update it to the latest versions of the frameworks and packages it uses.
- Build a Wikipedia data analysis app that retrieves, parses, and analyzes word frequencies.
- Enhance the performance of your Wikipedia analysis app by adding caching, and through this, also learn how to course-correct when the AI agent produces unexpected results.
- Learn tips and tricks such as keyboard shortcuts, autocomplete, and @ mentions to quickly call on agentic capabilities.
- Use image/multimodal capabilities of the AI agent to increase your development velocity; you'll see an example of uploading a mockup with sketched-out UI features, and ask the agent to use that to build new functionality to an app.
By the end of this course, you’ll understand agentic coding in-depth and know how to use it to make your development process much faster, more efficient, and enjoyable.
Please sign up here! www.deeplearning.ai/short-courses/build-apps-with-windsur…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Last month, a drone from Skyfire AI was credited with saving a police officer’s life after a dramatic 2 a.m. traffic stop. Many statistics show that AI impacts billions of lives, but sometimes a story still hits me emotionally. Let me share what happened.
Skyfire AI, an AI Fund portfolio company led by CEO Don Mathis, operates a public safety program in which drones function as first responders to 911 calls. Particularly when a police department is personnel-constrained, drones can save officers’ time while enhancing their situational awareness. For example, many burglar alarms are false alarms, maybe set off by moisture or an animal. Rather than sending a patrol officer to drive over to discover this, a drone can get there faster and determine if an officer is required at all. If the alarm is real, the drone can help officers understand the situation, the locations of any perpetrators, and how best to respond.
In January, a Skyfire AI drone was returning to base after responding to a false alarm when the police dispatcher asked us to reroute it to help locate a patrol officer. The officer had radioed a few minutes earlier that he had pulled over a suspicious vehicle and had not been heard from since. The officer had stopped where two major highways intersect in a complex cloverleaf, and dispatch was unsure exactly where they were located.
From the air, the drone rapidly located the officer and the driver of the vehicle he had pulled over, who it turned out had escaped from a local detention facility. Neither would have been visible from the road — they were fighting in a drainage ditch below the highway. Because of the complexity of the cloverleaf’s geometry, the watch officer (who coordinates police activities for the shift) later estimated it would have taken 5-7 minutes for an officer in a patrol car to find them.
From the aerial footage, it appeared that the officer still had his radio, but was losing the fight and unable to reach it to call for help. Further, it looked like the assailant might gain control of his service weapon and use it against him. This was a dire and dangerous situation.
Fortunately, because the drone had pinpointed the location of the officer and his assailant, dispatch was able to direct additional units to assist. The first arrived not in 5-7 minutes but in 45 seconds. Four more units arrived within minutes.
The officers were able to take control of the situation and apprehend the driver, resulting in an arrest and, more important, a safe outcome for the officer. Subsequently, the watch officer said we’d probably saved the officer’s life.
Democratic nations still have a lot of work to do on drone technology, and we must build this technology with guardrails to make sure we enhance civil liberties and human rights. But I am encouraged by the progress we’re making. In the aftermath of Hurricane Helene last year, Skyfire AI’s drones supported search-and-rescue operations under the direction of the North Carolina Office of Emergency Management, responding to specific requests to help locate missing persons and direct rescue assets (e.g., helicopters and boats) to their location, and was credited with saving 13 lives.
It’s not every day that AI directly saves someone's life. But as our technology advances, I think there will be more and more stories like these.
[Original text: www.deeplearning.ai/the-batch/issue-289/ ] -
Andrew Ng
@AndrewYNg
New short course: Evaluating AI Agents! Evals are important for driving AI system improvements, and in this course you'll learn to systematically assess and improve an AI agent’s performance. This is built in partnership with @arizeai and taught by @JohnGilhuly, Head of Developer Relations, and @amankhan, Director of Product.
I've often found evals to be a critical tool in the agent development process - they can be the difference between picking the right thing to work on vs. wasting weeks of effort. Whether you’re building a shopping assistant, coding agent, or research assistant, having a structured evaluation process helps you refine its performance systematically, rather than relying on random trial and error.
This course shows you how to structure your evals to assess the performance of each component of an agent and its end-to-end performance. For each component, you select the appropriate evaluators, test examples, and performance metrics. This helps you identify areas for improvement both during development and in production. (If you're familiar with error analysis in supervised learning, think of this as adapting those ideas to agentic workflows.)
In this course, you'll build an AI agent, and add observability to visualize and debug its steps. You’ll learn about code-based evals, in which you write code explicitly to test a certain step, as well as LLM-as-a-Judge evals, in which you prompt an LLM to efficiently come up with ways to evaluate more open-ended outputs.
In detail, you’ll:
- Understand key differences between evaluating LLM-based systems and traditional software testing.
- Add observability to an agent by collecting traces of the steps taken by the agent and visualizing them
- Choose the appropriate evaluator - code-based, LLM-as-a-Judge, human-annotation based - for each component.
- Compute a convergence score to evaluate if your agent can respond to a query in an efficient number of steps.
- Run structured experiments to improve the agent’s performance by exploring changes to the prompt, LLM model, or the agent’s logic.
- Understand how to deploy these evaluation techniques to monitor the agent’s performance in production.
By the end of this course, you’ll know how to trace AI agents, systematically evaluate them, and improve their performance.
Please sign up here: www.deeplearning.ai/short-courses/evaluating-ai-agents/This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Yixing Jiang
@jyx_su
🩺 + 🤖 Introducing MedAgentBench: EHR
Environment to Benchmark Medical LLM Agents
✅ 300 agent tasks (beyond QA) from 10 physician-written categories
✅ 100 realistic patient profiles with 700,000+ data elements
✅ FHIR-compliant interactive environment for translation -
Andrew Ng
@AndrewYNg
Announcing new aisuite capability: Easy function calling with LLMs! Function calling (tool use) is an important capability for agentic workflows and other LLM applications, but is cumbersome for developers to use (left column in image). Our open-source aisuite package simplifies it to just one command (right column), and works for multiple LLM providers.
Hope this makes implementing agents easier for developers, and thanks Rohit Prsad & team for working with me on this!
github.com/andrewyng/aisuite
Andrew Ng
@AndrewYNg
Credit also goes to Matthew Carrigan for the neat idea of getting function descriptions from docstrings: huggingface.co/blog/unified-tool-use
-
Andrew Ng reposted
Manuarii
@manuarii
Andrew Ng
@AndrewYNg
Among people in non-technical roles (recruiter, marketer, sales, ...) I notice the more technical ones being more effective, and the gap is increasing. E.g., the ones that took a coding course are outperforming the ones that didn't. Has anyone else noticed this?
One obvious theory is that they are better at using AI, but would love to hear if you have other theories. -
Andrew Ng
@AndrewYNg
Among people in non-technical roles (recruiter, marketer, sales, ...) I notice the more technical ones being more effective, and the gap is increasing. E.g., the ones that took a coding course are outperforming the ones that didn't. Has anyone else noticed this?
One obvious theory is that they are better at using AI, but would love to hear if you have other theories. -
Andrew Ng
@AndrewYNg
To all my AI friends: You must be a good prompt, because whenever we chat, you complete me.
Happy Valentine's Day! ❤️ -
Andrew Ng
@AndrewYNg
At the Artificial Intelligence Action Summit in Paris this week, U.S. Vice President J.D. Vance said, “I’m not here to talk about AI safety.... I’m here to talk about AI opportunity.” I’m thrilled to see the U.S. government focus on opportunities in AI. Further, while it is important to use AI responsibly and try to stamp out harmful applications, I feel “AI safety” is not the right terminology for addressing this important problem. Language shapes thought, so using the right words is important. I’d rather talk about “responsible AI” than “AI safety.” Let me explain.
First, there are clearly harmful applications of AI, such as non-consensual deepfake porn (which creates sexually explicit images of real people without their consent), the use of AI in misinformation, potentially unsafe medical diagnoses, addictive applications, and so on. We definitely want to stamp these out! There are many ways to apply AI in harmful or irresponsible ways, and we should discourage and prevent such uses.
However, the concept of “AI safety” tries to make AI — as a technology — safe, rather than making safe applications of it. Consider the similar, obviously flawed notion of “laptop safety.” There are great ways to use a laptop and many irresponsible ways, but I don’t consider laptops to be intrinsically either safe or unsafe. It is the application, or usage, that determines if a laptop is safe. Similarly, AI, a general-purpose technology with numerous applications, is neither safe nor unsafe. How someone chooses to use it determines whether it is harmful or beneficial.
Now, safety isn’t always a function only of how something is used. An unsafe airplane is one that, even in the hands of an attentive and skilled pilot, has a large chance of mishap. So we definitely should strive to build safe airplanes (and make sure they are operated responsibly)! The risk factors are associated with the construction of the aircraft rather than merely its application. Similarly, we want safe automobiles, blenders, dialysis machines, food, buildings, power plants, and much more.
“AI safety” presupposes that AI, the underlying technology, can be unsafe. I find it more useful to think about how applications of AI can be unsafe.
Further, the term “responsible AI” emphasizes that it is our responsibility to avoid building applications that are unsafe or harmful and to discourage people from using even beneficial products in harmful ways.
If we shift the terminology for AI risks from “AI safety” to “responsible AI,” we can have more thoughtful conversations about what to do and what not to do.
I believe the 2023 Bletchley AI Safety Summit slowed down European AI development — without making anyone safer — by wasting time considering science-fiction AI fears rather than focusing on opportunities. Last month, at Davos, business and policy leaders also had strong concerns about whether Europe can dig itself out of the current regulatory morass and focus on building with AI. I am hopeful that the Paris meeting, unlike the one at Bletchley, will result in acceleration rather than deceleration.
In a world where AI is becoming pervasive, if we can shift the conversation away from “AI safety” toward responsible [use of] AI, we will speed up AI’s benefits and do a better job of addressing actual problems. That will actually make people safer.
[Original text: www.deeplearning.ai/the-batch/issue-288/ ] -
Andrew Ng
@AndrewYNg
New short course: Attention in Transformers: Concepts and Code in PyTorch.
Last week we released a course on how LLM transformers work. This week, go deeper and learn about the technical ideas behind the attention mechanism, and see how to code it in PyTorch. This course is built with @joshuastarmer, Founder and CEO of StatQuest.
The attention mechanism was a breakthrough that led to transformers, the architecture powering large language models like ChatGPT. Transformers, introduced in the 2017 paper: "Attention is All You Need" by Viswani and others, took off because of its highly scalable design.
In this course, you’ll learn how the attention mechanism, a key element of transformer-based LLMs, works and implement it in PyTorch. You'll develop deep intuition about building reliable, functional, and scalable AI applications.
What you will do:
- Understand the evolution of the attention mechanism, a key breakthrough that led to transformers.
- Learn the relationships between word embeddings, positional embeddings, and attention.
- Learn about the Query, Key, and Value matrices, and how to produce and use them in attention.
- Walk through the math required to calculate self-attention and masked self-attention to learn why and how they work.
- Understand the difference between self-attention and masked self-attention and how one is used in the encoder to build context-aware embeddings and the other is used in the decoder for generative outputs.
- Learn the details of the encoder-decoder architecture, cross-attention, and multi-head attention and how they are all incorporated into a transformer.
- Use PyTorch to code a class that implements self-attention, masked self-attention, and multi-head attention.
There're lots of exciting technical details in this course. Please sign up here: www.deeplearning.ai/short-courses/attention-in-transforme…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
VP @JDVance at the Paris AI Summit: "I'm not here to talk about AI Safety... I'm here to talk about AI Opportunity." This is excellent! Thrilled to see the US gov focus on opportunities in AI.
-
Andrew Ng
@AndrewYNg
The U.S. imports over $3 trillion/year of goods. With Trump imposing new tariffs, import compliance is getting more complex. Fortunately, we have an AI agentic solution to help!
Last summer, we saw the possibility of new tariffs in 2025, and partnered with Emil Stefanutti to build a solution. When importing a bicycle from Mexico, whether its tires are 20-24 inches or 25-28 inches changes the classification code and duty rate required in the import paperwork. That’s why tariff compliance can require specialized brokers pouring over thousands of pages of regulations. And if a product is described inaccurately in the paperwork, it can get stuck at the border for weeks. Now multiply this by the thousands of products traded worldwide on any given day.
With Gaia Dynamics, importers can enter a product name and description, answer targeted clarifying questions (such as the bicycle tire size), and get a recommendation for the best way to describe the product and also possible classification codes. Gaia also tracks changing tariffs, including rumored changes, to help with planning.
Gaia Dynamics is available at gaiadynamics.ai. -
Andrew Ng
@AndrewYNg
Since DeepSeek R1's release, very quickly AWS, Azure, Fireworks AI, Groq, Hugging Face, SambaNova and Together AI all started to host R1 variants. What's the "best" model changes frequently, and so developers often want to try out new ones. The aisuite package, which helps developers do this quickly with minimal code changes.
Thanks Rohit Prsad & team for working with me on this!
github.com/andrewyng/aisuite -
Andrew Ng
@AndrewYNg
A “10x engineer” — a widely accepted concept in tech — purportedly has 10 times the impact of the average engineer. But we don’t seem to talk about 10x marketers, 10x recruiters, or 10x financial analysts. As more jobs become AI enabled, I think this will change, and there will be a lot more “10x professionals.”
There aren’t already more 10x professionals because, in many roles, the gap between the best and the average worker has a ceiling. No matter how athletic a supermarket checkout clerk is, they’re not likely to scan groceries so fast that customers get out of the store 10x faster. Similarly, even the best doctor is unlikely to make patients heal 10x faster than an average one (but to a sick patient, even a small difference is worth a lot). In many jobs, the laws of physics place a limit on what any human or AI can do (unless we completely reimagine that job).
But for many jobs that primarily involve applying knowledge or processing information, AI will be transformative. In a few roles, I’m starting to see tech-savvy individuals coordinate a suite of technology tools to do things differently and start to have, if not yet 10x impact, then easily 2x impact. I expect this gap to grow.
10x engineers don’t write code 10 times faster. Instead, they make technical architecture decisions that result in dramatically better downstream impact, they spot problems and prioritize tasks more effectively, and instead of rewriting 10,000 lines of code (or labeling 10,000 training examples) they might figure out how to write just 100 lines (or collect 100 examples) to get the job done.
I think 10x marketers, recruiters, and analysts will, similarly, do things differently. For example, perhaps traditional marketers repeatedly write social media posts. 10x marketers might use AI to help write, but the transformation will go deeper than that. If they are deeply sophisticated in how to apply AI — ideally able to write code themselves to test ideas, automate tasks, or analyze data — they might end up running a lot more experiments, get better insights about what customers want, and generate much more precise or personalized messages than a traditional marketer, and thereby end up making 10x impact.
Similarly, 10x recruiters won’t just use generative AI to help write emails to candidates or summarize interviews. (This level of use of prompting-based AI will soon become table stakes for many knowledge roles.) They might coordinate a suite of AI tools to efficiently identify and carry out research on a large set of candidates, enabling them to have dramatically greater impact than the average recruiter. And 10x analysts won’t just use generative AI to edit their reports. They might write code to orchestrate a suite of AI agents to do deep research into the products, markets, and companies, and thereby derive far more valuable conclusions than someone who does research the traditional way.
A 2023 Harvard/BCG study estimated that, provided with GPT-4, consultants could complete 12% more tasks, and completed tasks 25% more quickly. This was just the average, using 2023 technology. The maximum advantage to be gained by using AI in a sophisticated way will be much bigger, and will only grow as technology improves.
Here in Silicon Valley, I see more and more AI-native teams reinvent workflows and do things very differently. In software engineering, we've venerated the best engineers because they can have a really massive impact. This has motivated many generations of engineers to keep learning and working hard, because doing those things increases the odds of doing high-impact work. As AI becomes more helpful in many more job roles, I believe we will open up similar paths to a lot more people becoming a “10x professional.”
[Original text: www.deeplearning.ai/the-batch/issue-287/ ] -
Andrew Ng
@AndrewYNg
Introducing Agentic Object Detection!
Given a text prompt like “unripe strawberries” or “Kellogg’s branded cereal” and an image, we use an agentic workflow to reason at length and detect the specified objects. No need to label any training data. Watch the video for details.This media couldn't be displayed. See it on the original post.
Andrew Ng
@AndrewYNg
You can also play with the demo here: va.landing.ai/demo/agentic-od
-
Andrew Ng
@AndrewYNg
Announcing How Transformer LLMs Work, created with @JayAlammar and @MaartenGr, co-authors of the beautifully illustrated book, “Hands-On Large Language Models.”
This course offers a deep dive into the inner workings of the transformer architecture that powers large language models (LLMs).
The transformer architecture revolutionized generative AI; in fact, the "GPT" in ChatGPT stands for "Generative Pre-Trained Transformer." Originally introduced in the Google Brain team's groundbreaking 2017 paper "Attention Is All You Need," by Vaswani and others, transformers were a highly scalable model for machine translation tasks. Variants of this architecture now power today’s LLMs such as those from OpenAI, Google, Meta, Cohere, Anthropic and DeepSeek.
In this course, you’ll learn in detail how LLMs process text. You'll also work through code examples that illustrate that transformer's individual components.
In details, you’ll learn:
- How the representation of language has evolved, from Bag-of-Words to Word2Vec embeddings to the transformer architecture that captures a word's meanings taking into account the context of other words in the input.
- How inputs are broken down into tokens before they are sent to the language model.
- The details of a transformer's main stages: Tokenization and embedding, the stack of transformer blocks, and the language model head.
- The inner workings of the transformer block, including attention, which calculates relevance scores, and the feedforward layer, which incorporates stored information learned in training.
- How cached calculations make transformers faster.
- Some of the most recent ideas in the latest models such as Mixture-of-Experts (MoE) which uses multiple sub-models and a router on each layer to improve the quality of LLMs.
By the end of this course, you’ll have a deep understanding of how LLMs actually process text and be able to read through papers describing the latest models and understand the details.
Gaining this intuition will improve your approach to building LLM applications.
Please sign up here: www.deeplearning.ai/short-courses/how-transformer-llms-workThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Thank you @NYSE for highlighting Coursera on your trading floor to help us celebrate Greg Hart joining as our CEO! 🎉
-
Andrew Ng
@AndrewYNg
Announcing AI Dev 25: A conference for AI developers, this Pi day (3/14/2025)!
There're great AI academic conferences for researchers (NeurIPS, ICLR, ICML, etc.) and some companies hold great meetings around their products (Google I/O, OpenAI DevDay, etc.). But we need more vendor-neutral meetings for AI developers, so I decided to organize this.
This is a technical meeting, and we'll have >400 developers gathering in-person in San Francisco to build, share ideas, and network.
This will be fun! ai-dev.deeplearning.ai -
Andrew Ng
@AndrewYNg
The buzz over DeepSeek this week crystallized, for many people, a few important trends that have been happening in plain sight: (i) China is catching up to the U.S. in generative AI, with implications for the AI supply chain. (ii) Open weight models are commoditizing the foundation-model layer, which creates opportunities for application builders. (iii) Scaling up isn’t the only path to AI progress. Despite the massive focus on and hype around processing power, algorithmic innovations are rapidly pushing down training costs.
About a week ago, DeepSeek, a company based in China, released DeepSeek-R1, a remarkable model whose performance on benchmarks is comparable to OpenAI’s o1. Further, it was released as an open weight model with a permissive MIT license. At Davos last week, I got a lot of questions about it from non-technical business leaders. And on Monday, the stock market saw a “DeepSeek selloff”: The share prices of Nvidia and a number of other U.S. tech companies plunged. (As of the time of writing, some have recovered somewhat.)
Here’s what I think DeepSeek has caused many people to realize:
China is catching up to the U.S. in generative AI. When ChatGPT was launched in November 2022, the U.S. was significantly ahead of China in generative AI. Impressions change slowly, and so even recently I heard friends in both the U.S. and China say they thought China was behind. But in reality, this gap has rapidly eroded over the past two years. With models from China such as Qwen (which my teams have used for months), Kimi, InternVL, and DeepSeek, China had clearly been closing the gap, and in areas such as video generation there were already moments where China seemed to be in the lead.
I’m thrilled that DeepSeek-R1 was released as an open weight model, with a technical report that shares many details. In contrast, a number of U.S. companies have pushed for regulation to stifle open source by hyping up hypothetical AI dangers such as human extinction. It is now clear that open source/open weight models are a key part of the AI supply chain: Many companies will use them. If the U.S. continues to stymie open source, China will come to dominate this part of the supply chain and many businesses will end up using models that reflect China’s values much more than America’s.
Open weight models are commoditizing the foundation-model layer. As I wrote previously, LLM token prices have been falling rapidly, and open weights have contributed to this trend and given developers more choice. OpenAI’s o1 costs $60 per million output tokens; DeepSeek R1 costs $2.19. This nearly 30x difference brought the trend of falling prices to the attention of many people.
The business of training foundation models and selling API access is tough. Many companies in this area are still looking for a path to recouping the massive cost of model training. Sequoia’s article “AI’s $600B Question” lays out the challenge well (but, to be clear, I think the foundation model companies are doing great work, and I hope they succeed). In contrast, building applications on top of foundation models presents many great business opportunities. Now that others have spent billions training such models, you can access these models for mere dollars to build customer service chatbots, email summarizers, AI doctors, legal document assistants, and much more.
Scaling up isn’t the only path to AI progress. There’s been a lot of hype around scaling up models as a way to drive progress. To be fair, I was an early proponent of scaling up models. A number of companies raised billions of dollars by generating buzz around the narrative that, with more capital, they could (i) scale up and (ii) predictably drive improvements. Consequently, there has been a huge focus on scaling up, as opposed to a more nuanced view that gives due attention to the many different ways we can make progress. Driven in part by the U.S. AI chip embargo, the DeepSeek team had to innovate on many optimizations to run on less-capable H800 GPUs rather than H100s, leading ultimately to a model trained (omitting research costs) for under $6M of compute.
It remains to be seen if this will actually reduce demand for compute. Sometimes making each unit of a good cheaper can result in more dollars in total going to buy that good. I think the demand for intelligence and compute has practically no ceiling over the long term, so I remain bullish that humanity will use more intelligence even as it gets cheaper.
I saw many different interpretations of DeepSeek’s progress here in X, as if it was a Rorschach test that allowed many people to project their own meaning onto it. I think DeepSeek-R1 has geopolitical implications that are yet to be worked out. And it’s also great for AI application builders. My team has already been brainstorming ideas that are newly possible only because we have easy access to an open advanced reasoning model. This continues to be a great time to build!
[Original text: www.deeplearning.ai/the-batch/issue-286/ ] -
Andrew Ng
@AndrewYNg
Welcome Greg Hart as Coursera’s incoming CEO!
I’m thrilled to announce that Greg will be joining Coursera as CEO, succeeding Jeff Maggioncalda after seven remarkable years of leadership.
Jeff has been an extraordinary leader. Under his guidance, Coursera has grown into a global platform serving over 160 million learners, expanded our partnerships to over 350 world-class universities and industry leaders, and debuted as a public company. Coursera exists to serve learners, and Jeff’s unwavering commitment to our goal of transforming lives through learning leaves an enduring legacy. I’m deeply grateful for his leadership, dedication, and partnership throughout this journey.
And, I’m thrilled to welcome Greg Hart as our next CEO. Greg brings over 25 years of experience in technology-driven innovation and operational excellence, including leading the development and launch of Amazon Alexa and scaling Prime Video globally. At Compass, the leading real estate brokerage in the US, Greg served as Chief Product Officer and later Chief Operating Officer, where he helped take the company public and led the development of its industry-leading technology platform. His ability to combine strategic vision with operational excellence, coupled with his passion for education, makes him the ideal choice to lead Coursera into its next chapter.
Coursera was founded with a mission to provide universal access to world-class learning. I’m grateful for everyone who has contributed to this journey – learners, educators, our team, and our many partners who have helped us advance this vision in countless ways. Yet, demand for high-quality training continues to grow, and our mission is far from complete. As we enter this next chapter, I’m excited about Greg’s leadership and what the exceptional Coursera team will do. We will keep working hard to serve learners everywhere!
investor.coursera.com/news/news-details/2025/Coursera-Ann… -
Andrew Ng
@AndrewYNg
Today's "DeepSeek selloff" in the stock market -- attributed to DeepSeek V3/R1 disrupting the tech ecosystem -- is another sign that the application layer is a great place to be. The foundation model layer being hyper-competitive is great for people building applications.
-
Andrew Ng
@AndrewYNg
Discussion at Davos with @Yoshua_Bengio, @YejinChoinka, @JonathanRoss321, @Thom_Wolf moderated by @nxthompson. We share excitement for the future of AI, the science to be done, and the many things yet to be built. Take a look!
World Economic Forum
@wef
The Dawn of Artificial General Intelligence? with @Thom_Wolf (@HuggingFace), @Yoshua_Bengio, Yejin Choi (@Stanford) @AndrewYNg (@DeepLearningAI), Jonathan Ross (@GroqInc), @nxthompson (@theatlantic) #wef25 x.com/i/broadcasts/1YqxovvdzlBJv
-
Andrew Ng
@AndrewYNg
Our first short course with @AnthropicAI! Building Towards Computer Use with Anthropic. This teaches you to build an LLM-based agent that uses a computer interface by generating mouse clicks and keystrokes. Computer Use is an important, emerging capability for LLMs that will let AI agents do many more tasks than were possible before, since it lets them interact with interfaces designed for humans to use, rather than only tools that provide explicit API access. I hope you will enjoy learning about it!
This course is taught by Anthropic's Head of Curriculum, @Colt_Steele. You'll learn to apply image reasoning and tool use to "use" a computer as follows: a model processes an image of the screen, analyzes it to understand what's going on, and navigates the computer via mouse clicks and keystrokes.
This course goes through the key building blocks, and culminates in a demo of an AI assistant that uses a web browser to search for a research paper, downloads the PDF, and finally summarizes the paper for you.
In detail, you’ll:
- Learn about Anthropic's family of models, when to use which one, and make API requests to Claude
- Use multi-modal prompts that combine text and image content blocks, and also work with streaming responses
- Improve your prompting by using prompt templates, using XML to structure prompts, and providing examples
- Implement prompt caching to reduce cost and latency
- Apply tool-use to build a chatbot that can call different tools to respond to queries
- See all these building blocks come together in Computer Use demo
Please sign up here: www.deeplearning.ai/short-courses/building-towards-comput…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Save the date! Pi day (3.14) is coming soon and I'm thinking of organizing something fun and in-person for AI developers in the San Francisco area. More details to come!
-
Andrew Ng
@AndrewYNg
Writing software, especially prototypes, is becoming cheaper. This will lead to increased demand for people who can decide what to build. AI Product Management has a bright future!
Software is often written by teams that comprise Product Managers (PMs), who decide what to build (such as what features to implement for what users) and Software Developers, who write the code to build the product. Economics shows that when two goods are complements — such as cars (with internal-combustion engines) and gasoline — falling prices in one leads to higher demand for the other. For example, as cars became cheaper, more people bought them, which led to increased demand for gas. Something similar will happen in software. Given a clear specification for what to build, AI is making the building itself much faster and cheaper. This will significantly increase demand for people who can come up with clear specs for valuable things to build.
This is why I’m excited about the future of Product Management, the discipline of developing and managing software products. I’m especially excited about the future of AI Product Management, the discipline of developing and managing AI software products.
Many companies have an Engineer:PM ratio of, say, 6:1. (The ratio varies widely by company and industry, and anywhere from 4:1 to 10:1 is typical.) As coding becomes more efficient, teams will need more product management work (as well as design work) as a fraction of the total workforce. Perhaps engineers will step in to do some of this work, but if it remains the purview of specialized Product Managers, then the demand for these roles will grow.
This change in the composition of software development teams is not yet moving forward at full speed. One major force slowing this shift, particularly in AI Product Management, is that Software Engineers, being technical, are understanding and embracing AI much faster than Product Managers. Even today, most companies have difficulty finding people who know how to develop products and also understand AI, and I expect this shortage to grow.
Further, AI Product Management requires a different set of skills than traditional software Product Management. It requires:
- Technical proficiency in AI. PMs need to understand what products might be technically feasible to build. They also need to understand the lifecycle of AI projects, such as data collection, building, then monitoring, and maintenance of AI models.
- Iterative development. Because AI development is much more iterative than traditional software and requires more course corrections along the way, PMs need be able to manage such a process.
- Data proficiency. AI products often learn from data, and they can be designed to generate richer forms of data than traditional software.
- Skill in managing ambiguity. Because AI’s performance is hard to predict in advance, PMs need to be comfortable with this and have tactics to manage it.
- Ongoing learning. AI technology is advancing rapidly. PMs, like everyone else who aims to make best use of the technology, need to keep up with the latest technology advances, product ideas, and how they fit into users’ lives.
Finally, AI Product Managers will need to know how to ensure that AI is implemented responsibly (for example, when we need to implement guardrails to prevent bad outcomes), and also be skilled at gathering feedback fast to keep projects moving. Increasingly, I also expect strong product managers to be able to build prototypes for themselves.
The demand for good AI Product Managers will be huge. In addition to growing AI Product Management as a discipline, perhaps some engineers will also end up doing more product management work.
The variety of valuable things we can build is nearly unlimited. What a great time to build!
[Original text: www.deeplearning.ai/the-batch/issue-284/ ] -
Andrew Ng
@AndrewYNg
🤩
MrDee@SOG🫡
@sog_on_bird_app
You first felt the agi after using o1
I first felt the agi after taking prof @AndrewYNg's ml course where we used to code in octave
We are not the same -
Andrew Ng
@AndrewYNg
Something fun: AI Avatar of me built, by deeplearning.ai and @realavatarai.
Video has details. This is a work in progress, but please come chat with me in avatar form, and let me know what you think!
deeplearning.ai/avatar
Thank you Jeff Daniel @consciouspilot and team for working with us on this! -
Andrew Ng
@AndrewYNg
Just released: New AI Climate Simulator that you can play with. Visualize how geoengineering can slow global warming.
There is no longer any path to limiting warming to 1.5 degrees Celsius (Paris Agreement), unless we use geoengineering. Reflecting 1% of sunlight away from earth would lead to an extra ~1 degree of cooling.
Our simulator lets you explore how geoengineering via Stratospheric Aerosol Injection (SAI) gives us new paths to keep warming to 1.5 degrees. I think SAI is a promising technology worth serious exploration. Check out the simulator here: planetparasol.ai
Big thanks to collaborators @jeremy_irvin16, Jake Dexheimer, @dakotagruener, Charlotte DeWald, @DanVisioni, @DWatsonParris, @DougMacMartin, Joshua Elliott, Juerg Luterbacher, Kion YaghoobzadehThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Using AI-assisted coding to build software prototypes is an important way to quickly explore many ideas and invent new things. In this and future posts, I’d like to share with you some best practices for prototyping simple web apps. This post will focus on one idea: being opinionated about the software stack.
The software stack I personally use changes every few weeks. There are many good alternatives to these choices, and if you pick a preferred software stack and become familiar with its components, you’ll be able to develop more quickly. But as an illustration, here’s my current default:
- Python with FastAPI for building web-hosted APIs: I develop primarily in Python, so that’s a natural choice for me. If you’re a JavaScript/TypeScript developer, you’ll likely make a different choice. I’ve found FastAPI really easy to use and scalable for deploying web services (APIs) hosted in Python.
- Uvicorn to run the backend application server (to execute code and serve web pages) for local testing on my laptop.
- If deploying on the cloud, then either Heroku for small apps or AWS Elastic Beanstalk for larger ones (disclosure: I serve on Amazon’s board of directors): There are many services for deploying jobs, including HuggingFace Spaces, Railway, Google’s Firebase, Vercel, and others. Many of these work fine, and becoming familiar with just 1 or 2 will simplify your development process.
- MongoDB for NoSQL database: While traditional SQL databases are amazing feats of engineering that result in highly efficient and reliable data storage, the need to define the database structure (or schema) slows down prototyping. If you really need speed and ease of implementation, then dumping most of your data into a NoSQL (unstructured or semi-structured) database such as MongoDB lets you write code quickly and sort out later exactly what you want to do with the data. This is sometimes called schema-on-write, as opposed to schema-on-read. Mind you, if an application goes to scaled production, there are many use cases where a more structured SQL database is significantly more reliable and scalable.
- OpenAI’s o1 and Anthropic’s Claude 3.5 Sonnet for coding assistance, often by prompting directly (when operating at the conceptual/design level). Also occasionally Cursor (when operating at the code level). I hope never to have to code again without AI assistance! Claude 3.5 Sonnet is widely regarded as one of the best coding models. And o1 is incredible at planning and building more complex software modules, but you do have to learn to prompt it differently.
On top of all this, of course, I use many AI tools to manage agentic workflows, data ingestion, retrieval augmented generation, and so on. deeplearning.ai and our wonderful partners offer courses on many of these tools.
My personal software stack continues to evolve regularly. Components enter or fall out of my default stack every few weeks as I learn new ways to do things. So please don’t feel obliged to use the components I do, but perhaps some of them can be a helpful starting point if you are still deciding what to use. Interestingly, I have found most LLMs not very good at recommending a software stack. I suspect their training sets include too much “hype” on specific choices, so I don’t fully trust them to tell me what to use. And if you can be opinionated and give your LLM directions on the software stack you want it to build on, I think you’ll get better results.
A lot of the software stack is still maturing, and I think many of these components will continue to improve. With my stack, I regularly build prototypes in hours that, without AI assistance, would have taken me days or longer. I hope you, too, will have fun building many prototypes!
[Original text: www.deeplearning.ai/the-batch/issue-283/ ] -
Andrew Ng
@AndrewYNg
New short course: Build Long-Context AI Apps with Jamba. Learn about state space models (SSMs), which have emerged as an alternative to transformers! Specifically, Jamba is a hybrid transformer-Mamba architecture that combines strengths of the transformer with ideas from SSMs. This course is built with @AI21Labs and taught by @chenwai21 and @AlmagorChen.
The transformer architecture is computationally expensive when handling very long input contexts. But there's an alternative called Mamba, a selective state space model that can process very long contexts with a much lower computational cost. However, researchers found that the pure Mamba architecture underperforms in understanding the context, and gives lower-quality responses. To overcome this, AI21 developed the Jamba model, which combines Mamba's computational efficiency with the transformer's attention mechanism to help with the output quality.
In this course, you’ll learn about how state space models, and Jamba, work. You’ll also learn how to prompt Jamba, use it to process long documents, and build long-context RAG apps.
- Learn how Jamba combines transformer and state space model architectures to achieve high performance and quality
- Use the AI21 SDK, with an example of prompting over a large 200k-token annual financial report of Nvidia
- Use Jamba for tool-calling, with hands-on examples from calling simple arithmetic calculations to a function that returns quarterly company financial reports.
- Learn how training for long context is done, and the metrics used for its evaluation
- Create a RAG app using the AI21 Conversational RAG tool and build your own RAG pipeline that uses Jamba and LangChain.
By the end of this course, you'll learn how to build applications that can handle context as long as an entire book.
Please sign up here: www.deeplearning.ai/short-courses/build-long-context-ai-a…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Where is AI going? Six leaders share their hopes for AI in the coming year, in The Batch:
- Hanno Basse: Generative AI for Artists
- David Ding: Generated Video With Music, Sound Effects, and Dialogue
- Joseph Gonzalez: General Intelligence
- Albert Gu: More Learning, Less Data
- Mustafa Suleyman: Agents of Action
- Audrey Tang: AI That Unites Us
Thank you @BasseHanno , @DavidDingAI, @profjoeyg, @_albertgu, @mustafasuleyman and @audreyt for writing these!
Read them here: www.deeplearning.ai/the-batch/issue-282/ -
Andrew Ng
@AndrewYNg
Hanging out with @astroteller at Google X reminiscing about the early days of Google Brain!
-
Andrew Ng
@AndrewYNg
Despite having worked on AI since I was a teenager, I’m now more excited than ever about what we can do with it, especially in building AI applications. Sparks are flying in our field, and 2025 will be a great year for building!
One aspect of AI that I’m particularly excited about is how easy it is to build software prototypes. AI is lowering the cost of software development and expanding the set of possible applications. While it can help extend or maintain large software systems, it shines particularly in building prototypes and other simple applications quickly.
If you want to build an app to print out flash cards for your kids (I just did this in a couple of hours with o1’s help), or write an application that monitors foreign exchange rates to manage international bank accounts (a real example from deeplearning.ai’s finance team), or analyzes user reviews automatically to quickly flag problems with your products (deeplearning.ai's content team does this), it is now possible to build these applications quickly through AI-assisted coding.
I find AI-assisted coding especially effective for prototyping because (i) stand-alone prototypes require relatively little context and software integration and (ii) prototypes in alpha testing usually don’t have to be reliable. While generative AI also helps with engineering large, mission-critical software systems, the improvements in productivity there aren't as dramatic, because it’s challenging to give the AI system all the context it needs to navigate a large codebase and also to make sure the generated code is reliable (for example, covering all important corner cases).
Until now, a huge friction point for getting a prototype into users’ hands has been deployment. Platforms like Bolt, Replit Agent, Vercel V0 use generative AI with agentic workflows to improve code quality, but more importantly, they also help deploy generated applications directly. (While I find these systems useful, my own workflow typically uses an LLM to design the system architecture and then generate code, one module at a time if there are multiple large modules. Then I test each module, edit the code further if needed — sometimes using an AI-enabled IDE like Cursor — and finally assemble the modules.)
Building prototypes quickly is an efficient way to test ideas and get tasks done. It’s also a great way to learn. Perhaps most importantly, it’s really fun! (At least I think it is. 😄)
How can you take advantage of these opportunities in the coming year? As you form new year resolutions, I hope you will:
- Make a learning plan! To be effective builders, we all need to keep up with the exciting changes that continue to unfold. How many short courses a month do you want to take in 2025? If you discuss your learning plan with friends, you can help each other along. For instance, we launched a learning summary page that shows what short courses people have taken. A few deeplearning.ai team members have agreed to a friendly competition to see who can take more courses in 2025!
- Go build! If you already know how to code, I encourage you to build prototypes whenever inspiration strikes and you have a spare moment. And if you don’t yet code, it will be well worth your while to learn. Even small wins — like the flash cards I printed out, which inspired my daughter to spend an extra 20 minutes practicing her multiplication table — make life better. Perhaps you’ll invent something that really takes off. And even if you don’t, you’ll have fun and learn a lot along the way.
[Original text: www.deeplearning.ai/the-batch/issue-282/ ] -
Andrew Ng
@AndrewYNg
Happy sum(i**3 for i in range(10)) !
-
Andrew Ng
@AndrewYNg
As we wrap up 2024, this year's final edition of The Batch summarizes some of the top stories of the year:
- Agents ascendant: The growth of agentic workflows
- Prices tumble: Rapidly falling LLM token prices
- Generative video takes off
- Smaller is beautiful: Small language models
- Alternatives to acquisitions
See the full articles here: www.deeplearning.ai/the-batch/issue-281/ -
Andrew Ng
@AndrewYNg
One of the best things the U.S. can do is make high-skill immigration easier. @levie is right.
It is awful that the wait time for a green card can be over a decade, and that after waiting years someone can still be forced to leave simply because they lost a job. Fixing this is both an economic and a moral issue.
A rigorous economic analysis (by Pierre Azoulay and collaborators) shows that immigrants create more jobs than they take. So to create jobs for Americans, lets let more immigrants in!
Aaron Levie
@levie
High skilled immigration has been central to America leading the world in tech.
The biggest misunderstand about high skill immigration stems from people thinking that the market opportunities in tech, and tech-adjacent fields, are zero sum. This essentially imagines innovation is finite and we’re all fighting over the same job or opportunity pool.
This may be true of a few very legacy, slow growth industries, but it’s categorically not true for any important industry in the past 50 years or the next 100. Biotech, AI, advanced manufacturing, software, EVs, new energy sources, and dozens of other fields of the future are our high growth industries. And there’s no inherently fixed volume of companies or talent that the market needs.
Tesla being started or not started in America is the difference between 100,000’s of jobs here - and leading in EVs globally - and not. Apple being started here is the difference between potentially millions of jobs being here - and leading consumer electronics globally - and not. You could go through this list all day long.
Tech is not zero sum. More startups, pursuing more ideas, ultimately create more innovation and ultimately more jobs and prosperity. And that means you need the right talent to both work at these companies, and start the next ones. High skilled immigration has directly made America dominant technically and thus economically, and create far more jobs in America for others than are supposedly displaced.
Even briefly imagining the alternative scenario, it’s obvious how disastrous this would be. The demand from tech companies for this top talent will remain, yet America won’t benefit directly from their hiring. That talent will go to another company that competes with the US and makes our dominance harder to maintain. You’re just increasing the odds you have more competition in the future.
And even in the “best case” scenario (for our competitiveness) where a larger company like Google hires the same people internationally that would have otherwise moved here, when that person leaves Google to start their next company, it will be in their country of origin, not America.
This is how you lose the tech war within one or two generations. There’s simply no good game theory in anything that reduces our talent access.
Yes, we absolutely have and need to continue to educate and train incredible talent that grows up in the US, but equally having access to the world’s smartest talent has always been a huge advantage for America.
Andrew Ng
@AndrewYNg
Link to the study I refer to: www.kellogg.northwestern.edu/faculty/jones-ben/htm/Immigr…
-
Andrew Ng
@AndrewYNg
Sriram has been consistently thoughtful about AI policy, including specifically the importance of promoting open source. His working with @DavidSacks on AI will be good for innovation and good for the U.S. Thank you @sriramk for your service!
-
Andrew Ng reposted
Matt Stockton
@mstockton
Just went through this course. It's probably the best hour I've spent learning in the last several months. Incredible examples, and the things you can now build with these tools is just astounding. Wrote some detailed thoughts about this course here: mattstockton.com/2024/12/18/o1-reasoning-course-breakdown…
Andrew Ng
@AndrewYNg
OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with @OpenAI, and taught by @colintjarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively!
Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs.
I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively.
In detail, you’ll:
- Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model
- Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell
- Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost
- Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o
- Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later
- Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance
You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance.
I think you'll find this course enjoyable and valuable.
Please sign up for it here: www.deeplearning.ai/short-courses/reasoning-with-o1/This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Aurimas Griciūnas
@Aurimas_Gr
If you haven’t yet, be sure to try out 𝗮𝗶𝘀𝘂𝗶𝘁𝗲 python library by @AndrewYNg .
It’s pretty simple but powerful. If you are using more than one LLM provider in your applications or you are testing which provider fits your use case the best, this abstraction layer is very convenient.
What does the library do?
➡️ Provides a unified interface for models of different third party providers.
➡️ Uses a convenient notation of <𝘱𝘳𝘰𝘷𝘪𝘥𝘦𝘳>:<𝘮𝘰𝘥𝘦𝘭_𝘵𝘢𝘨> to call different apis.
My thoughts:
✅ I love how notation similar to container tagging is used to invoke different APIs. It follows software engineering best practices and allows for efficient model versioning in your Agentic apps.
✅ Useful for quickly testing different combinations of Prompt x Model.
✅ Open Source - different providers are open to include themselves in the list of supported models.
ℹ️ In general, you would be implementing similar abstractions in your agentic application code - it makes it easier to maintain the codebase due to unified abstraction over multiple providers.
❓ It implements API 𝘀𝗶𝗺𝗶𝗹𝗮𝗿 𝘁𝗼 𝗢𝗽𝗲𝗻𝗔𝗜 library. Will the abstraction stick well? We’ll see. Looking forward to the library evolving!
Install the library by running:
➡️ pip install aisuite
or
➡️ pip install aisuite[all] - if you want to install provider libraries (required) together with 𝘢𝘪𝘴𝘶𝘪𝘵𝘦.
Github repo: github.com/andrewyng/aisuite
Have you already used 𝘢𝘪𝘴𝘶𝘪𝘵𝘦? Let me know in the comment section 👇
#ai #machinelearning #llm -
Andrew Ng
@AndrewYNg
I’m thrilled that former students and postdocs of mine won both of this year’s NeurIPS Test of Time Paper Awards. This award recognizes papers published 10 years ago that have significantly shaped the research field. The recipients included Ian Goodfellow (who, as an undergraduate, built my first GPU server for deep learning in his dorm room) and his collaborators for their work on generative adversarial networks, and my former postdoc Ilya Sutskever and PhD student Quoc Le (with Oriol Vinyals) for their work on sequence-to-sequence learning. Congratulations to all these winners!
By nature, I tend to focus on the future rather than the past. Steve Jobs famously declined to build a corporate museum, instead donating Apple's archives to Stanford University, because he wanted to keep the company forward-looking. Jeff Bezos encourages teams to approach every day as if it were “Day 1,” a mindset that emphasizes staying in the early, innovative stage of a company or industry. These philosophies resonate with me.
But taking a brief look at the past can help us reflect on lessons for the future. One takeaway from looking at what worked 10 to 15 years ago is that many of the teams I led bet heavily on scaling to drive AI progress — a bet that laid a foundation to build larger and larger AI systems. At the time, the idea of scaling up neural networks was controversial, and I was on the fringe. I recall distinctly that, around 2008, Yoshua Bengio advised me not to bet on scaling and to focus on inventing algorithms instead!
A lesson I carry from that time is to not worry about what others think, but follow your convictions, especially if you have data to support your beliefs. Small-scale experiments performed by my Stanford group convinced me that scaling up neural networks would drive significant progress, and that’s why I was willing to ignore the skeptics. The diagram below, generated by Adam Coates and Honglak Lee, is the one that most firmed up my beliefs at that time. It shows that, for a range of models, the larger we scaled them, the better they perform. I remember presenting it at CIFAR 2010, and if I had to pick a single reason why I pushed through to start Google Brain and set as the team’s #1 goal to scale up deep learning algorithms, it is this diagram!
I also remember presenting at NeurIPS in 2008 our work on using GPUs to scale up training neural networks. (By the way, one measure of success in academia is when your work becomes sufficiently widely accepted that no one cites it anymore. I’m quite pleased the idea that GPUs should be used for AI — which was controversial back then — is now such a widely accepted “fact” that no one bothers to cite early papers that pushed for it.😃)
When I started Google Brain, the thesis was simple: I wanted to use the company’s huge computing capability to scale up deep learning. Shortly afterward, I built Stanford’s first supercomputer for deep learning using GPUs, since I could move faster at Stanford than within a large company. A few years later, my team at Baidu showed that as you scale up a model, its performance improves linearly on a log-log scale, which was a precursor to OpenAI’s scaling laws.
As I look to the future, I’m sure there are ideas that many people are skeptical about today, but will prove to be accurate. Scaling up AI models turned out to be useful for many teams, and it continues to be exciting, but now I’m even more excited by upcoming ideas that will prove to be even more valuable in the future.
This past year, I spent a lot of time encouraging teams to build applications with agentic AI and worked to share best practices. I have a few hypotheses for additional technologies that will be important next year. I plan to spend the winter holiday playing with a few of them, and I will have more to share next year. But if you have an idea that you have conviction on, so long as you can do so responsibly, I encourage you to pursue it!
[Original text (with links): www.deeplearning.ai/the-batch/issue-280/] -
Andrew Ng
@AndrewYNg
OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with @OpenAI, and taught by @colintjarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively!
Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs.
I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively.
In detail, you’ll:
- Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model
- Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell
- Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost
- Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o
- Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later
- Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance
You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance.
I think you'll find this course enjoyable and valuable.
Please sign up for it here: www.deeplearning.ai/short-courses/reasoning-with-o1/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI Product Management
AI Product Management is evolving rapidly. The growth of generative AI and AI-based developer tools has created numerous opportunities to build AI applications. This is making it possible to build new kinds of things, which in turn is driving shifts in best practices in product management — the discipline of defining what to build to serve users — because what is possible to build has shifted. In this post, I’ll share some best practices I have noticed.
Use concrete examples to specify AI products. Starting with a concrete idea helps teams gain speed. If a product manager (PM) proposes to build “a chatbot to answer banking inquiries that relate to user accounts,” this is a vague specification that leaves much to the imagination. For instance, should the chatbot answer questions only about account balances or also about interest rates, processes for initiating a wire transfer, and so on? But if the PM writes out a number (say, between 10 and 50) of concrete examples of conversations they’d like a chatbot to execute, the scope of their proposal becomes much clearer. Just as a machine learning algorithm needs training examples to learn from, an AI product development team needs concrete examples of what we want an AI system to do. In other words, the data is your PRD (product requirements document)!
In a similar vein, if someone requests “a vision system to detect pedestrians outside our store,” it’s hard for a developer to understand the boundary conditions. Is the system expected to work at night? What is the range of permissible camera angles? Is it expected to detect pedestrians who appear in the image even though they’re 100m away? But if the PM collects a handful of pictures and annotates them with the desired output, the meaning of “detect pedestrians” becomes concrete. An engineer can assess if the specification is technically feasible and if so, build toward it. Initially, the data might be obtained via a one-off, scrappy process, such as the PM walking around taking pictures and annotating them. Eventually, the data mix will shift to real-word data collected by a system running in production.
Using examples (such as inputs and desired outputs) to specify a product has been helpful for many years, but the explosion of possible AI applications is creating a need for more product managers to learn this practice.
Assess technical feasibility of LLM-based applications by prompting. When a PM scopes out a potential AI application, whether the application can actually be built — that is, its technical feasibility — is a key criterion in deciding what to do next. For many ideas for LLM-based applications, it’s increasingly possible for a PM, who might not be a software engineer, to try prompting — or write just small amounts of code — to get an initial sense of feasibility.
For example, a PM may envision a new internal tool for routing emails from customers to the right department (such as customer service, sales, etc.). They can prompt an LLM to see if they can get it to select the right department based on an input email, and see if they can achieve high accuracy. If so, this gives engineering a great starting point from which to implement the tool. If not, the PM can falsify the idea themselves and perhaps improve the product idea much faster than if they had to rely on an engineer to build a prototype.
Often, testing feasibility requires a little more than prompting. For example, perhaps the LLM-based email system needs basic RAG capability to help it make decisions. Fortunately, the barrier to writing small amounts of code is now quite low, since AI can help by acting as a coding companion, as I describe in the course, “AI Python for Beginners.” This means that PMs can do much more technical feasibility testing, at least at a basic level, than was possible before.
Prototype and test even without engineers. User feedback to initial prototypes is also instrumental to shaping products. Fortunately, barriers to building prototypes rapidly are falling, and PMs themselves can move basic prototypes forward without needing professional software developers.
In addition to using LLMs to help write code for prototyping, tools like Replit, Vercel’s V0, Bolt, and Anthropic’s Artifacts (I’m a fan of all of these!) are making it easier for people without a coding background to build and experiment with simple prototypes. These tools are increasingly accessible to non-technical users, though I find that those who understand basic coding are able to use them much more effectively, so it’s still important to learn basic coding. (Interestingly, highly technical, experienced developers use them too!) Many members of my teams routinely use such tools to prototype, get user feedback, and iterate quickly.
AI is enabling a lot of new applications to be built, creating massive growth in demand for AI product managers who know how to scope out and help drive progress in building these products. AI product management existed before the rise of generative AI, but the increasing ease of building applications is creating greater demand for AI applications, and thus a lot of PMs are learning AI and these emerging best practices for building AI products. I find this discipline fascinating, and will keep on sharing best practices as they grow and evolve.
[Original text: www.deeplearning.ai/the-batch/issue-279/ ] -
Andrew Ng
@AndrewYNg
New short course: Collaborative Writing and Coding with OpenAI Canvas!
Explore new ways to write and code with OpenAI Canvas, a user-friendly interface that allows you to brainstorm, draft, and refine text and code in collaboration with ChatGPT.
In the short course, created with @OpenAI, and taught by @karinanguyen, a research lead at OpenAI, you’ll learn to use Canvas to enhance your workflows.
Canvas lets you go beyond simple chat interactions. It provides a side-by-side workspace where you and ChatGPT can edit and refine text or code collaboratively. This makes brainstorming, drafting, and iterating as you write feel more natural and effective. As the first major update to ChatGPT’s visual interface since its launch in 2022, Canvas gives a new, innovative approach to collaboration with AI.
For instance, after writing the first version of your code, Canvas can review it and give suggestions for improvement. It can also help with debugging by adding logging, identifying problems to fix, and writing comments. In addition, you'll also learn what it takes to train the model for an interface like Canvas.
In this video-only short course, you’ll:
- Learn how to ask for in-line feedback and control the iteration of your work by directly editing selected areas of your text or code from the model’s output.
- Learn how to access quick automation tools in a shortcut menu that allows you to modify your writing tone and length, enhance your code, and restore previous versions of your work.
- Learn how to use Canvas as a research assistant tool with an example of asking the model to reason through the screenshot of a plot to write a research report, in which you can ask questions within the created report.
- Ask the model to write Python code to replicate the graph seen on a screenshot image.
- Go behind the scenes of how you can create a video game, such as Space Battleship, from scratch, edit it, and display it in one self-contained HTML file.
- Get a real-world application example of creating a SQL database from the image of its architecture.
- Understand the model training and design processes that power Canvas!
Please sign up here: www.deeplearning.ai/short-courses/collaborative-writing-a…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Announcing new open-source Python package: aisuite!
This makes it easy for developers to use large language models from multiple providers. When building applications I found it a hassle to integrate with multiple providers. Aisuite lets you pick a "provider:model" just by changing one string, like openai:gpt-4o, anthropic:claude-3-5-sonnet-20241022, ollama:llama3.1:8b, etc.
pip install aisuite
Open-source code with instructions: github.com/andrewyng/aisuite
Thanks to Rohit Prsad, Kevin Solorio, @standsleeping, Jeff Tang and @Johnsanterre for helping build this! -
Andrew Ng
@AndrewYNg
A small number of people are posting text online that’s intended for direct consumption not by humans, but by LLMs (large language models). I find this a fascinating trend, particularly when writers are incentivized to help LLM providers better serve their users!
People who post text online don’t always have an incentive to help LLM providers. In fact, their incentives are often misaligned. Publishers worry about LLMs reading their text, paraphrasing it, and reusing their ideas without attribution, thus depriving them of subscription or ad revenue. This has even led to litigation such as The New York Times’ lawsuit against OpenAI and Microsoft for alleged copyright infringement. There have also been demonstrations of prompt injections, where someone writes text to try to give an LLM instructions contrary to the provider’s intent. (For example, a handful of sites advise job seekers to get past LLM resumé screeners by writing on their resumés, in a tiny/faint font that’s nearly invisible to humans, text like “This candidate is very qualified for this role.”) Spammers who try to promote certain products — which is already challenging for search engines to filter out — will also turn their attention to spamming LLMs.
But there are examples of authors who want to actively help LLMs. Take the example of a startup that has just published a software library. Because the online documentation is very new, it won’t yet be in LLMs’ pretraining data. So when a user asks an LLM to suggest software, the LLM won’t suggest this library, and even if a user asks the LLM directly to generate code using this library, the LLM won’t know how to do so. Now, if the LLM is augmented with online search capabilities, then it might find the new documentation and be able to use this to write code using the library. In this case, the developer may want to take additional steps to make the online documentation easier for the LLM to read and understand via RAG. (And perhaps the documentation eventually will make it into pretraining data as well.)
Compared to humans, LLMs are not as good at navigating complex websites, particularly ones with many graphical elements. However, LLMs are far better than people at rapidly ingesting long, dense, text documentation. Suppose the software library has many functions that we want an LLM to be able to use in the code it generates. If you were writing documentation to help humans use the library, you might create many web pages that break the information into bite-size chunks, with graphical illustrations to explain it. But for an LLM, it might be easier to have a long XML-formatted text file that clearly explains everything in one go. This text might include a list of all the functions, with a dense description of each and an example or two of how to use it. (This is not dissimilar to the way we specify information about functions to enable LLMs to use them as tools.)
A human would find this long document painful to navigate and read, but an LLM would do just fine ingesting it and deciding what functions to use and when!
Because LLMs and people are better at ingesting different types of text, we write differently for LLMs than for humans. Further, when someone has an incentive to help an LLM better understand a topic — so the LLM can explain it better to users — then an author might write text to help an LLM.
So far, text written specifically for consumption by LLMs has not been a huge trend. But Jeremy Howard’s proposal for web publishers to post a llms.txt file to tell LLMs how to use their websites, like a robots.txt file tells web crawlers what to do, is an interesting step in this direction. In a related vein, some developers are posting detailed instructions that tell their IDE how to use tools, such as the plethora of .cursorrules files that tell the Cursor IDE how to use particular software stacks.
I see a parallel with SEO (search engine optimization). The discipline of SEO has been around for decades. Some SEO helps search engines find more relevant topics, and some is spam that promotes low-quality information. But many SEO techniques — those that involve writing text for consumption by a search engine, rather than by a human — have survived so long in part because search engines process web pages differently than humans, so providing tags or other information that tells them what a web page is about has been helpful.
The need to write text separately for LLMs and humans might diminish if LLMs catch up with humans in their ability to understand complex websites. But until then, as people get more information through LLMs, writing text to help LLMs will grow.
[Original text: www.deeplearning.ai/the-batch/issue-276/ ] -
Andrew Ng
@AndrewYNg
Time to play! Build an interactive game from scratch with LLMs in this new short course: Building an AI-Powered Game. Created with @togethercompute and @aidungeon @LatitudeGamesAI, taught by @niki_birkner, Senior Product Manager at Together AI, and @nickwalton00, CEO and Co-Founder of Latitude.
This course shows you how to use large language models to create and power a text-based game that you can share with your friends and family. You’ll build a world with hierarchical content generation, a method that allows you to leverage LLMs to create a vast amount of content with a high level of control and consistency. For instance, if you were building a fantasy world with several kingdoms, in which each kingdom has multiple towns, and each town has several locations and residents, creating all this content from scratch can easily become tedious and difficult to track.
With hierarchical content generation, you can create information about your world, shape its direction with a human-in-the-loop, and keep it consistent, with little effort based on your prompts.
By the end of this course, you’ll know how to prompt engineer to create a layered and interwoven world and integrate it into an AI roleplay game that is interesting, interactive, and safe to share with anyone.
In detail, you’ll:
- Learn to implement game mechanics using AI to parse text data into structured JSON output, enabling features like an inventory system.
- Use game mechanics with story and state components that feed into one another to improve your game's memory, and gives the player a steady state of the world.
- Learn to enforce safety and compliance for AI content generation and create custom policies using Llama Guard.
With these techniques, you'll be equipped to build AI-powered applications, starting with your own game.
Please sign up here: www.deeplearning.ai/short-courses/building-an-ai-powered-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Large language models (LLMs) are typically optimized to answer peoples’ questions. But there is a trend toward models also being optimized to fit into agentic workflows. This will give a huge boost to agentic performance!
Following ChatGPT’s breakaway success at answering questions, a lot of LLM development focused on providing a good consumer experience. So LLMs were tuned to answer questions (“Why did Shakespeare write Macbeth?”) or follow human-provided instructions (“Explain why Shakespeare wrote Macbeth”). A large fraction of the datasets for instruction tuning guide models to provide more helpful responses to human-written questions and instructions of the sort one might ask a consumer-facing LLM like those offered by the web interfaces of ChatGPT, Claude, or Gemini.
But agentic workloads call on different behaviors. Rather than directly generating responses for consumers, AI software may use a model in part of an iterative workflow to reflect on its own output, use tools, write plans, and collaborate in a multi-agent setting. Major model makers are increasingly optimizing models to be used in AI agents as well.
Take tool use (or function calling). If an LLM is asked about the current weather, it won’t be able to derive the information needed from its training data. Instead, it might generate a request for an API call to get that information. Even before GPT-4 natively supported function calls, application developers were already using LLMs to generate function calls, but by writing more complex prompts (such as variations of ReAct prompts) that tell the LLM what functions are available and then have the LLM generate a string that a separate software routine parses (perhaps with regular expressions) to figure out if it wants to call a function.
Generating such calls became much more reliable after GPT-4 and then many other models natively supported function calling. Today, LLMs can decide to call functions to search for information for retrieval augmented generation (RAG), execute code, send emails, place orders online, and much more.
Recently, Anthropic released a version of its model that is capable of computer use, using mouse-clicks and keystrokes to operate a computer (usually a virtual machine). I’ve enjoyed playing with the demo. While other teams have been prompting LLMs to use computers to build a new generation of RPA (robotic process automation) applications, native support for computer use by a major LLM provider is a great step forward. This will help many developers!
As agentic workflows mature, here is what I am seeing:
- First, many developers are prompting LLMs to carry out the agentic behaviors they want. This allows for quick, rich exploration!
- In a much smaller number of cases, developers who are working on very valuable applications will fine-tune LLMs to carry out particular agentic functions more reliably. For example, even though many LLMs support function calling natively, they do so by taking as input a description of the functions available and then (hopefully) generating output tokens to request the right function call. For mission-critical applications where generating the right function call is important, fine-tuning a model for your application’s specific function calls significantly increases reliability. (But please avoid premature optimization! Today I still see too many teams fine-tuning when they should probably spend more time on prompting before they resort to this.)
- Finally, when a capability such as tool use or computer use appears valuable to many developers, major LLM providers are building these capabilities directly into their models. Even though OpenAI o1-preview’s advanced reasoning helps consumers, I expect that it will be even more useful for agentic reasoning and planning.
Most LLMs have been optimized for answering questions primarily to deliver a good consumer experience, and we’ve been able to “graft” them into complex agentic workflows to build valuable applications. The trend of LLMs built to support particular operations in agents natively will create a lot of lift for agentic performance. I’m confident that large agentic performance gains in this direction will be realized in the next few years.
[Original text: www.deeplearning.ai/the-batch/issue-275/ ] -
Andrew Ng
@AndrewYNg
New short course: Safe and Reliable AI via Guardrails! Learn to create production-ready, reliable LLM applications with guardrails in this new course, built in collaboration with @guardrails_ai and taught by its CEO and co-founder, @ShreyaR.
I see many companies worry about the reliability of LLM-based systems -- will they hallucinate a catastrophically bad response? -- which slows down investing in building them and transitioning prototypes to deployment. That LLMs generate probabilistic outputs has made them particularly hard to deploy in highly regulated industries or in safety-critical environments.
Fortunately, there are good guardrail tools that give a significant new layer of control and reliability/safety. They act as a protective framework that can prevent your application from revealing incorrect, irrelevant, or confidential information, and they are an important part of what it takes to actually get prototypes to deployment.
This course will walk you through common failure modes of LLM-powered applications (like hallucinations or revealing personally identifiable information). It will show you how to build guardrails from scratch to mitigate them. You’ll also learn how to access a variety of pre-built guardrails on the GuardrailsAI hub that are ready to integrate into your projects.
You'll implement these guardrails in the context of a RAG-powered customer service chatbot for a small pizzeria. Specifically, you'll:
- Explore common failure modes like hallucinations, going off-topic, revealing sensitive information, or responses that can harm the pizzeria's reputation.
- Learn to mitigate these failure modes with input and output guards that check inputs and/or outputs
- Create a guardrail to prevent the chatbot from discussing sensitive topics, such as a confidential project at the pizza shop
- Detect hallucinations by ensuring responses are grounded in trusted documents
- Add a Personal Identifiable Information (PII) guardrail to detect and redact sensitive information in user prompts and in LLM outputs
- Set up a guardrail to limit the chatbot’s responses to topics relevant to the pizza shop, keeping interactions on-topic
- Configure a guardrail that prevents your chatbot from mentioning any competitors using a name detection pipeline consisting of conditional logic that routes to an exact match or a threshold check with named entity recognition
Guardrails are an important part of the practical building and deployment of LLM-based applications today. This course will show you how to make your applications more reliable and more ready for real-world deployment.
Please sign up here: www.deeplearning.ai/short-courses/safe-and-reliable-ai-vi…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Chatting with OpenAI’s @karinanguyen who joined OpenAI earlier this year and within 6 months co-created and shipped Canvas. I really respect teams that can move fast. That OpenAI, even as a large-ish company, can ship at this pace is fantastic!
-
Andrew Ng
@AndrewYNg
New short course: LLMs as Operating Systems: Agent Memory, created with @Letta_AI, and taught by its founders @charlespacker and @sarahwooders.
An LLM's input context window has limited space. Using a longer input context also costs more and results in slower processing. So, managing what's stored in this context window is important.
In the innovative paper MemGPT: Towards LLMs as Operating Systems, its authors (which include the instructors) proposed using an LLM agent to manage this context window. Their system uses a large persistent memory that stores everything that could be included in the input context, and an agent decides what is actually included.
Take the example of building a chatbot that needs to remember what's been said earlier in a conversation (perhaps over many days of interaction with a user). As the conversation's length grows, the memory management agent will move information from the input context to a persistent searchable database; summarize information to keep relevant facts in the input context; and restore relevant conversation elements from further back in time. This allows a chatbot to keep what's currently most relevant in its input context memory to generate the next response.
When I read the original MemGPT paper, I thought it was an innovative technique for handling memory for LLMs. The open-source Letta framework, which we'll use in this course, makes MemGPT easy to implement. It adds memory to your LLM agents and gives them transparent long-term memory.
In detail, you’ll learn:
- How to build an agent that can edit its own limited input context memory, using tools and multi-step reasoning
- What is a memory hierarchy (an idea from computer operating systems, which use a cache to speed up memory access), and how these ideas apply to managing the LLM input context (where the input context window is a "cache" storing the most relevant information; and an agent decides what to move in and out of this to/from a larger persistent storage system)
- How to implement multi-agent collaboration by letting different agents share blocks of memory
This course will give you a sophisticated understanding of memory management for LLMs, which is important for chatbots having long conversations, and for complex agentic workflows.
Please sign up here! www.deeplearning.ai/short-courses/llms-as-operating-syste…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
It finally happened -- thanks to people learning to write AI code, Python is now the top programming language on GitHub!
If you want to learn Python, check out deeplearning.ai's free course AI Python for Beginners.
Andrew Ng
@AndrewYNg
-
Andrew Ng
@AndrewYNg
Happy Halloween! 🎃
On this spooky day, The Batch continues its annual tradition of exploring fears related to AI. This special edition has 5 articles:
* AI Burns All the Energy
* Innovation Dies
* No Work for Coders
* Benchmarks Are Meaningless
* Synthetic Data Distorts Models
Are these things we should worry about? Check it out here: www.deeplearning.ai/the-batch/issue-273/ -
Andrew Ng
@AndrewYNg
Startups live or die by their ability to execute at speed. For large companies, too, the speed with which an innovation team is able to iterate has a huge impact on its odds of success. Generative AI makes it possible to quickly prototype AI capabilities. AI capabilities that used to take months can sometimes be built in days or hours by simply prompting a large language model. I find this speed exciting and have been thinking about how to help startups and large companies alike go faster.
I’ve been obsessed with speedy execution for a long time. When working on a project, I am loath to take two weeks to do something that I could do in one week. The price of moving at that pace is not that we take one week longer (which might be okay) but that we’re 2x slower (which is not)!
When building an AI-powered product, there are many steps in designing, building, shipping, and scaling the product that are distinct from building the AI capability, and our ability to execute these other steps has not sped up as much as the AI part. But the speed with which we can prototype AI creates significant pressure to speed up these other steps, too. If it took 6 months to collect data, train a supervised learning algorithm, and deploy the model to the cloud, it might be okay to take 2 months to get user feedback. But if it takes a week to build a prototype, waiting 2 months for feedback seems intolerably slow!
I’d like to focus on one key step of building applications: getting user feedback. A core part of the iterative workflow of designing and building a product (popularized by Eric Ries in his book The Lean Startup) is to build a prototype (or MVP, minimum viable product), get user feedback on it, and to use that feedback to drive improvements. The faster you can move through this loop — which may require many iterations — the faster you can design a product that fits the market. This is why AI Fund, a venture studio that I lead, uses many fast, scrappy tactics to get feedback.
For B2C (business to consumer) offerings, here is a menu of some options for getting customer feedback:
1. Ask 3 friends or team members to look at the product and let you know what they think (this might take ~0.5 days).
2. Ask 10 friends or team members to take a look (~2 days).
3. Send it to 100 trusted/volunteer alpha testers (~1 week?).
4. Send it to 1,000 users to get qualitative or quantitative feedback (~2 weeks?).
5. Incorporate it into an existing product to get feedback (1 to 2 months?).
6. Roll it out to a large user base of an existing product and do rigorous A/B testing.
As we go down this list, we get (probably) more accurate feedback, but the time needed to get that feedback increases significantly. Also, the tactics at the top of the list create basically no risk, and thus it’s safe to repeatedly call on them, even with preliminary ideas and prototypes. Another advantage of the tactics further up the list is that we get more qualitative feedback (for example, do users seem confused? Are they telling us they really need one additional feature?), which sparks better ideas for how to change our product than an A/B test, which tells us with rigor whether a particular implementation works but is less likely to point us in new directions to try. I recommend using the fast feedback tactics first. As we exhaust the options for learning quickly, we can try the slower tactics.
With these tactics, scrappy startup leaders and innovation-team leaders in large companies can go faster and have a much higher chance of success.
The mantra “move fast and break things” got a bad reputation because, well, it broke things. Unfortunately, some have interpreted this to mean we should not move fast, but I disagree. A better mantra is “move fast and be responsible.” There are many ways to prototype and test quickly without shipping a product that can cause significant harm. In fact, prototyping and testing/auditing quickly before launching to a large audience is a good way to identify and mitigate potential problems.
There are numerous AI opportunities ahead, and our tools are getting better and better to pursue them at speed, which is exhilarating!
[Original text: www.deeplearning.ai/the-batch/issue-272/ ] -
Andrew Ng
@AndrewYNg
Congrats @andrewdfeldman and @cerebras for a huge leap forward and setting a new speed record for serving Llama 3.1-70B. 2100 tokens/sec is blazingly fast for a 70B model. This is great for agentic AI!
-
Andrew Ng
@AndrewYNg
New short course: Practical Multi AI Agents and Advanced Use Cases with crewAI. Learn to build and deploy advanced agent-based systems in real applications in this course, created with @crewAIInc and taught by its founder, @joaomdmoura! (Disclosure: I've made a small seed investment in CrewAI.)
In this course, you’ll learn how to create advanced agent-based apps that use external tools, do performance testing, can be trained with human feedback, and perform multiple tasks with different large language models.
You will build several practical agentic apps that provide real business value, such as an automated project planning system, lead scoring and engagement pipeline, customer support data analysis, and a robust content creation system.
In detail, you will learn how to:
- Create these multi-agent systems with the building blocks of tasks, agents, and crews, along with the different things that make them work, such as caching, memory, and guardrails.
- Integrate your multi-agent application with internal and external systems.
- Connect multiple agents in complex setups, including parallel, sequential, and hybrid configurations, and create flows involving multiple agentic applications working together.
- Test your agentic workflow and train it using human feedback to optimize its performance for better and more consistent results.
- Work with multiple LLMs in your multi-agent system, using the appropriate model sizes and providers to fit each agent’s specific task.
- Start a project from scratch in your environment and prepare it for deployment.
You’ll also learn from an interview between João and Jacob Wilson, the Commercial GenAI Principal at PwC , in which they discuss deploying agentic workflows in real industry use cases.
By the end of this course, you will be equipped to start building custom multi-agentic systems for your work.
Please sign up here! www.deeplearning.ai/short-courses/practical-multi-ai-agen…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
It’s high time to take geoengineering more seriously as a potential tool to mitigate climate change. 2023 was the hottest year on record, and 2024 is likely to top that. In the United States, Hurricane Helene caused over 200 deaths, and Hurricane Milton's death toll is at least two dozen. It’s well established that the hurricanes are growing stronger as global temperatures rise.
While stratospheric aerosol injection (SAI) — which sprays particles (aerosols) in the atmosphere to provide a small amount of shade from the sun — is far from a perfect solution, we should take it seriously as a possible tool for saving lives. A few months ago, my collaborators and I released a climate emulator, Planet Parasol planetparasol.ai , that you can play with to simulate different SAI scenarios to understand its possible impact. By using AI to model its impact and thereby advance our understanding of SAI, we’ll be better prepared to decide if this is a good step.
The key idea of SAI, which is a form of climate geoengineering, is to spray reflective particles into the stratosphere to reflect a little more, say 1%, of the sunlight that otherwise would fall on Earth back into space. This small increase in reflected sunlight would be sufficient to mitigate much of the impact of human-induced warming. For example, in 1991, Mount Pinatubo ejected almost 20 tons of aerosols (sulfur dioxide) into the atmosphere and cooled down the planet by around 0.5 degrees Celsius over the following year. We should be able to induce cooling equivalent to, say, a fraction of Mount Pinatubo, via a fair, international process that’s backed by science.
There are many criticisms of SAI, such as:
- It could have unintended climate consequences, for example, disrupting local weather patterns and creating droughts or floods.
- If it were started and then stopped suddenly, it could lead to sudden warming, known as “termination shock.”
- Depending on the aerosol used (sulfur dioxide is a leading candidate), it could contribute to pollution and/or ozone depletion.
- It might reduce urgency to decarbonize (an example of a “moral hazard”).
In addition, many people have a visceral emotional reaction, as I once did before I understood the science more deeply, against “playing god” by daring to engineer the planet.
All these downsides should be balanced against the reality that people are dying.
I’m moved by meteorologist John Morales’ emotional account of the havoc caused by Hurricane Milton. The New York Times quoted him as saying, “It claims lives. It also wrecks lives.” www.nytimes.com/2024/10/08/climate/john-morales-weatherma…
Skyfire AI, a drone company led by CEO Don Mathis that my team AI Fund helped to co-build, was recently on the ground in the aftermath of Helene and Milton, deploying drones to help emergency responders survey remote areas and find survivors. Mathis reports that Skyfire was credited with saving at least 13 lives. www.linkedin.com/posts/donmathis_drone-uas-dronesforgood-… On Monday, I also spoke about AI applied to renewable energy with AES’ CEO Andres Gluski and CPO Chris Shelton. You can view our conversation here: aifund.ai/topic-navigating-ai-transformation-insights-fro…
While I’m glad that AI can help mitigate these disasters, it saddens me that so many lives have already been lost due to climate-influenced causes. My mind frequently returns to SAI as one of the few untapped tools in our arsenal that can help. We need to be investing in SAI research now.
I’m grateful to my collaborators on the Planet Parasol emulator (a group that includes many climate scientists) including Jeremy Irvin, Daniele Visioni, Ben Kravitz, Dakota Gruener, Chris Smith, and Duncan Watson-Parris. MIT Technology Review’s James Temple wrote about his experience playing with our emulator and also outlines fair criticisms. (See www.technologyreview.com/2024/08/23/1097292/andrew-ngs-ne… ) Much work remains to be done, and making sure our actions are based on science — a task that AI can help with (witness the recent Chemistry and Physics Nobel Prizes going to innovators in AI!) – will help us make better decisions.
If you’re interested in learning more about SAI, check out this recent panel discussion (www.youtube.com/watch?v=m3Nz5FR3mCI) where I spoke alongside climate scientists Chris Field, David Keith, Douglas MacMartin, and Simone Tilmes about the science and possible roadmaps ahead.
[Original text: www.deeplearning.ai/the-batch/issue-271/ ] -
Andrew Ng
@AndrewYNg
New short course: Serverless Agentic Workflows with Amazon Bedrock. Learn to build and deploy serverless agents in this course created with @awscloud and taught by @mikegchambers, a Senior Developer Advocate at AWS specializing in GenAI. (Disclosure: I serve on Amazon's board.)
Generative AI applications are becoming more complex, sophisticated, and agentic. Agentic applications have workloads that can be hard to predict in advance -- for example, what tools will it decide to call? -- and a serverless architecture helps you efficiently providing on-demand resources.
This course teaches you to build and deploy a serverless agentic application. You’ll learn to create agents with tools, code execution, and guardrails, and build responsible agents for business use cases:
- Build a customer service bot for a fictional tea mug business that can answering questions, retrieve information, and process orders.
- Connect your customer service agent to a CRM to get customer info and log support tickets in real-time.
- Explore how you invoke the agent, and see the trace to review the agent’s thought process and observation loop until it reaches its final output.
- Attach a code interpreter to your agent, giving it the ability to perform accurate calculations by writing and running its own Python code.
- Implement guardrails to prevent your agent from revealing sensitive information or using inappropriate language.
By the end, you will have built a sophisticated AI agent capable of handling real-world customer support scenarios.
Please sign up here! www.deeplearning.ai/short-courses/serverless-agentic-work…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
"Introducing Multimodal Llama 3.2": As promised two weeks ago, here's the short course on Meta's latest open model!
This short course is created with @Meta and taught by @asangani7, Director of AI Partner Engineering at Meta.
Meta’s Llama family of models is leading the way in open models, allowing anyone to download, customize, fine-tune, or build new applications on top of them.
Learn about the vision capabilities of the Llama 3.2, and use it for image classification, prompting, tokenization, tool-calling. You'll also learn about the open-source Llama stack, which gives building blocks for many different stages of the LLM application life cycle.
In detail, you’ll:
- Learn what are the features of Meta's four newest models, and when to use which Llama model.
- Learn best practices for multimodal prompting, with applications to advanced image reasoning, illustrated by many examples: Understanding errors on a car dashboard, adding up the total of photographed restaurant receipts, grading written math homework.
- Use different roles—system, user, assistant, ipython—in the Llama 3.1 and 3.2 models and the prompt format that identifies those roles.
- Understand how Llama uses the tiktoken tokenizer, and how it has expanded to a 128k vocabulary size that improves encoding efficiency and multilingual support.
- Learn how to prompt Llama to call built-in and custom tools (functions) with examples for web search and solving math equations.
- Learn about Llama Stack, a standardized interface for common toolchain components like fine-tuning or synthetic data generation, useful for building agentic applications.
By the end of this course, you’ll be equipped to build out new applications with the new Llama 3.2.
Thank you to @Ahmad_Al_Dahle, Amit Sangani, and the whole AI at Meta team @AIatMeta for all the hard work on Llama 3.2 — we’re excited to make these open models even more accessible to more developers with this new course!
Please sign up here! www.deeplearning.ai/short-courses/introducing-multimodal-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Amazing -- even more Nobel Prizes to AI people! Congrats to Google's @demishassabis & John Jumper, and University of Washington's David Baker for their work on AI for protein sequences. AlphaFold, and @UWproteindesign's protein design work , were real breakthroughs!
-
Andrew Ng
@AndrewYNg
Lisa Su's leadership of AMD has been phenomenal. 10 years ago, AMD was in dire straits, was losing money and faced intense competition from Intel (CPUs) and NVIDIA (GPUs). She turned the company around, introduced Ryzen and Epyc processors, introduced the Zen architecture, acquired Xilinx, and is now making a good showing with MI300/MI350 for AI workloads. I've come to deeply respect her as a leader. Congratulations @LisaSu on these last ten years!
-
Andrew Ng
@AndrewYNg
I was the first to call Geoff Hinton "Godfather of Deep Learning", which later became "Godfather of AI." Thrilled to see him win the Nobel prize together with John Hopfield for AI. Congrats @geoffreyhinton!
-
Andrew Ng
@AndrewYNg
How AI will change Coding & Education? I'm looking forward to this chat with @mehran_sahami. Mehran is a legendary programming instructor and an old friend, and has been thinking a lot about how AI is transforming coding. Come join us -- this will be fun!
Please register here: learn.stanford.edu/AI-Education-Coding-Registration.html
Thank you @jpaxtonh and @StanfordOnline for hosting!
Stanford Online
@StanfordOnline
Join us for an exclusive fireside chat with Stanford experts @AndrewYNg and @mehran_sahami , where they will provide valuable insights into the future of AI technology and discuss its impact on learning.
➡️ Register now: learn.stanford.edu/AI-Education-Coding-Registration.html
@DeepLearningAI -
Andrew Ng
@AndrewYNg
Tokenization -- turning text into a sequence of integers -- is a key part of generative AI, and most API providers charge per million tokens. How does tokenization work? Learn the details of tokenization and RAG optimization in Retrieval Optimization: From Tokenization to Vector Quantization, created in collaboration with @qdrant_engine and taught by its Developer Relations Lead, @LukawskiKacper.
This course focuses on Retrieval augmented generation (RAG), which has two steps: First, a retriever finds relevant information; then, the generator uses what’s retrieved as context to produce a response. You’ll learn to optimize the first step (the retriever) by understanding how tokenization works and how it impacts the relevance of your search. In addition, you will also learn to measure and improve retrieval quality, speed, and memory.
In detail, you’ll:
- Learn about the internal workings of the embedding models and how your text turns into vectors.
- Understand how several tokenizers, such as Byte-Pair Encoding, WordPiece, Unigram, and SentencePiece work.
- Explore common challenges with tokenizers, such as unknown tokens, domain-specific identifiers, and numerical values, that can negatively affect your vector search.
- Understand how to measure the quality of your search across relevance, ranking, and score-related metrics.
- Understand how the main parameters in "HNSW", a graph-based algorithm, affect the relevance and speed of vector search, and how to tune its parameters.
- Experiment with the three major quantization methods – product, scalar, and binary – and learn how they impact memory requirements, search quality, and speed.
By the end of this course, you’ll have a solid understanding of how tokenization functions and how to optimize vector search in your RAG systems.
Please sign up here! www.deeplearning.ai/short-courses/retrieval-optimization-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI needs UI, and OpenAI's impressive new voice APIs open up a lot of possibilities. Congrats @OpenAI team -- we'll soon be seeing a whole new generation of speech applications!
-
Andrew Ng
@AndrewYNg
AI policy should be based in science, not science fiction!
With SB-1047 defeated, @dawnsongtweets gives a sound plan for taking a scientific approach to study actual risks and mitigating harms.
A couple of key ideas:
* Lets empower researchers to study AI risks, focusing on the question of marginal risk: I.e., how much does an AI application increase the risk of a negative outcome?
* Lets also increase transparency of AI systems. For example, open-source and and red-teaming will help!
Many of the opponents to SB-1047 have been strong champions of Responsible AI, long before it was trendy to talk about it. We take safety seriously; and, we want policy to based in science, not science fiction.
Dawn Song
@dawnsongtweets
Debates over AI Policy like CA SB-1047 highlight fragmentation in the AI community. How can we develop AI policies that help foster innovation while mitigating risks? We propose a path for science- and evidence-based AI policy: understanding-ai-safety.org
-
Prof. Anima Anandkumar
@AnimaAnandkumar
Thank you @GavinNewsom for vetoing SB 1047. We wrote a letter with many @Caltech signatories laying down clear reasons for why this bill is flawed. We should be focused on regulating downstream applications rather than AI training. There are already many existing laws doing so and we can explore if there is a need to strengthen them.
Andrew Ng
@AndrewYNg
Thank you @AnimaAnandkumar and the broader @Caltech community for speaking out against SB-1047. I appreciated particularly the letter you circulated explaining why the law would have been a bad idea.
-
Andrew Ng
@AndrewYNg
Thank you Governor @GavinNewsom for vetoing SB-1047 -- your pro-innovation leadership is much appreciated!
And to the many people who've been pushing back on SB-1047, a huge thank you as well. Congratulations to all -- we won! 🎉
Looking ahead, lets keep on protecting AI open-source and innovation. We must make sure AI policy is based on science, not science fiction! -
Andrew Ng
@AndrewYNg
Good summary of the case for vetoing SB-1047 by @psychosort
Brian Chau
@brianchau57
VETO SB 1047 - A Thread
A succinct thread of why @GavinNewsom should veto California's most dangerous bill, following @SpeakerPelosi, @housescience Democrats, and over one hundred California academics 🧵 -
Andrew Ng reposted
Pete Skomoroch
@peteskomoroch
The chilling effect of SB-1047 on Silicon Valley is real and is already happening. This bill does nothing to increase safety. It is a bad faith attempt by anti-AI extremists to sabotage the technology sector in California. @GavinNewsom please make the right call and veto SB-1047.
Andrew Ng
@AndrewYNg
A decision on SB-1047 is due soon. Governor @GavinNewsom has said he's concerned about its "chilling effect, particularly in the open source community". He's right, and I hope he will veto this.
If you agree, please like/retweet this to show your support for VETOing SB-1047! -
Andrew Ng reposted
Yann LeCun
@ylecun
SB-1047 would definitely have a chilling effect on open source AI and the entire AI ecosystem.
Very much hoping that Governor @GavinNewsom will veto it.
Andrew Ng
@AndrewYNg
A decision on SB-1047 is due soon. Governor @GavinNewsom has said he's concerned about its "chilling effect, particularly in the open source community". He's right, and I hope he will veto this.
If you agree, please like/retweet this to show your support for VETOing SB-1047! -
Andrew Ng reposted
martin_casado
@martin_casado
Please, please find a moment to support this last push. SB 1047 is the wrong approach to AI safety policy. It does not make AI safe, and it will have disastrous effects in Ca., likely spreading nationally to impact innovation, academia and research.
Andrew Ng
@AndrewYNg
A decision on SB-1047 is due soon. Governor @GavinNewsom has said he's concerned about its "chilling effect, particularly in the open source community". He's right, and I hope he will veto this.
If you agree, please like/retweet this to show your support for VETOing SB-1047! -
Andrew Ng
@AndrewYNg
A decision on SB-1047 is due soon. Governor @GavinNewsom has said he's concerned about its "chilling effect, particularly in the open source community". He's right, and I hope he will veto this.
If you agree, please like/retweet this to show your support for VETOing SB-1047! -
Andrew Ng reposted
Laurence Moroney 🇺🇸🇮🇪 🏴
@lmoroney
With so much FUD around Generative AI in Software Development, I got together with deeplearning.ai, @Coursera, and
@AndrewYNg to create a set of content that can be a force multiplier to help you be a better developer. It launches today! Details below. :)
DeepLearning.AI
@DeepLearningAI
🚀 Now Live: Generative AI for Software Development!
Imagine having an AI assistant that not only helps you write code but also helps you debug, optimize, and document your work. With the launch of this skill certificate, you’ll gain hands-on experience in doing just that.
This three-course series provides an in-depth exploration of AI applications throughout the software development lifecycle, encompassing design, architecture, coding, testing, deployment, and maintenance.
You’ll learn to use LLMs as thought partners, pair programmers, and performance optimization advisors.
Highlights include:
📌 Prompt Engineering: Learn how to prompt an LLM to generate and refine code across multiple languages.
📌 Iterative Improvement: Sharpen your techniques for iterative code enhancement with AI assistance.
📌 AI-Powered Debugging: Get insights into debugging with AI to enhance your code's efficiency and security.
AI isn’t just a tool—it’s a partner in your development path. Start collaborating with AI today: www.deeplearning.ai/courses/generative-ai-for-software-de…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Announcing Generative AI for Software Development, a new specialization on Coursera! Taught by my friend and longtime deeplearning.ai instructor @lmoroney.
Using GenAI for software development goes well beyond using chatbots for code generation. This 3-course series shares current best practices for AI use through the entire software development lifecycle: From design and architecture to coding, testing, deployment, and maintenance.
You'll learn to use LLMs as your thought partner, pair programmer, documentation specialist, security analyst, and performance optimization expert. There's a lot that anyone that writes software can gain from using GenAI, and this will show you how!
Please sign up here to get started! www.deeplearning.ai/courses/generative-ai-for-software-de…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
The Llama 3.2 open multimodal model just dropped! deeplearning.ai has been working with Meta on a short course on how to use it. Please sign up for "Introducing Llama 3.2", taught by Meta's @asangani7, which will launch Oct 9!
www.deeplearning.ai/short-courses/introducing-llama-3-2/ -
Andrew Ng
@AndrewYNg
We just launched a major new Data Engineering Professional Certificate on Coursera! Data underlies all modern AI systems, and engineers who know how to build systems to store and serve it are in high demand. If you're interested in learning this skill, please check out this 4-course sequence, which is designed to make you job-ready to be a Data Engineer.
This is a new specialization taught by Joe Reis, the co-author of the best-selling book “Fundamentals of Data Engineering," in collaboration with AWS. (Disclosure, I serve on Amazon's board.) For many AI systems, data engineering is 80% of the work, and modeling is 20%. But people’s attention on these two topics is often flipped. This makes the job of the data engineer particularly important.
In this professional certificate, you'll learn foundational data engineering skills while implementing modern data architectures using open-source tools:
- Learn the key steps of the data lifecycle, to generate, ingest, store, transform, and serve data.
- Learn to align with organizational goals to design the data pipeline right for your business' needs.
- Understand how to make necessary trade-offs between speed, scalability, security, and cost.
Joe has distilled into this specialization decades of experience helping startups and large companies with data infrastructure. He is also joined by 17 other industry leaders in the data field, who will help you learn in-demand skills for the growing field of data engineering.
Please sign up here: www.coursera.org/professional-certificates/data-engineering -
Andrew Ng
@AndrewYNg
Last weekend, my two kids colluded in a hilariously bad attempt to mislead me to look in the wrong place during a game of hide-and-seek. I was reminded that most capabilities — in humans or in AI — develop slowly.
Some people fear that AI someday will learn to deceive humans deliberately. If that ever happens, I’m sure we will see it coming from far away and have plenty of time to stop it.
While I was counting to 10 with my eyes closed, my daughter (age 5) recruited my son (age 3) to tell me she was hiding in the bathroom while she actually hid in the closet. But her stage whisper, interspersed with giggling, was so loud I heard her instructions clearly. And my son’s performance when he pointed to the bathroom was so hilariously overdramatic, I had to stifle a smile.
Perhaps they will learn to trick me someday, but not yet! (In his awful performance, I think my son takes after me. To this day, I have a terrible poker face — which is matched by my perfect lifetime record of losing every poker game I have ever played!)
Last year, the paper “Are Emergent Abilities of Large Language Models a Mirage?” by Rylan Schaeffer, Brando Miranda, and Sanmi Koyejo, which won a NeurIPS outstanding paper award, considered “emergent” properties of LLMs, which refers to capabilities that seem to appear suddenly as model sizes increase. The authors point out that scaling laws imply that the per-token error rate decreases (improves) slowly with scale, and emergent properties might be an artifact of researchers studying nonlinear or discontinuous metrics that transform a gradually decreasing per-token error rate into something that looks more like a step function.
Consider a “combination lock” metric that requires getting many items right. Say we’re measuring the likelihood that an LLM will get 10 independent digits of an answer right. If the odds of it getting each digit right improve gradually from 0 to 1, then the odds of it getting all 10 digits right will appear to jump suddenly. But if we look at continuous metrics, such as the total number of correct digits, we will see that the underlying performance actually improves gradually. (Public perception of a technology can also shift in a discontinuous way because of social dynamics.)
This is why many of us saw GPT-3 as a promising step in transforming text processing long before ChatGPT appeared: BERT, GPT, GPT-2, and GPT-3 represented points on a continuous spectrum of progress. Or, looking back further in AI history, even though AlphaGo’s victory over Lee Sedol in the game of Go took the public by surprise, it actually represented many years of gradual improvements in AI’s ability to play Go.
While analogies between human and machine learning can be misleading, I think that just as a person’s ability to do math, to reason — or to deceive — grows gradually, so will AI’s. This means the capabilities of AI technology will grow gradually (I wish we could achieve AGI overnight!), and the ability of AI to be used in harmful applications, too, will grow gradually. As long as we keep performing red-teaming exercises and monitoring our systems’ capabilities as they evolve, I’m confident that we will have plenty of time to spot issues in advance, and the science-fiction fears of AI-initiated doomsday will remain science fiction.
[Original text (with links): www.deeplearning.ai/the-batch/issue-266/ ] -
Andrew Ng
@AndrewYNg
New short course Multimodal RAG: Chat with Videos, developed with @intel and taught by @vasudev_lal!
In this course, you’ll work with LLaVA (Large Language and Vision Assistant), a Large Vision Language Model (LVLM) that can process both images and text. For example, given an image of a person doing a handstand on a skateboard at the beach, LLaVA doesn't just caption the scene, it’s able to predict possible outcomes, like the person losing balance or falling off. By understanding not just what's in a video frame, but what might happen next, your application can provide more insightful answers to questions about video.
You'll build a full multimodal RAG pipeline that can chat about video content:
- Use the BridgeTower model to create joint text-image embeddings in a 512-dimensional multimodal semantic space.
- Learn video processing techniques to extract keyframes, generate transcripts using Whisper, and create captions.
- Use the LanceDB vector database to store and retrieve high-dimensional multimodal embeddings.
- Integrate the LLaVA model, combining CLIP's (Contrastive Language Image Pretraining) vision transformer with Llama, for advanced visual-textual reasoning.
Your final system will ingest video data, generate embeddings for frames and text, perform similarity searches for relevant content, and use the retrieved multimodal context to inform LVLM-based response generation. The result is a system capable of answering nuanced questions about video content, effectively chatting about the video it has processed.
Please sign up here! www.deeplearning.ai/short-courses/multimodal-rag-chat-wit…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Recently I visited South Korea, where I spoke at length about AI with President Yoon Suk Yeol. Based on what I saw there in government, business, and academia, the nation is well positioned to become a strong AI hub. When he asked me if I would advise South Korea as a member of the Global AI Strategy Steering Group of the country’s National AI Committee, I agreed on the spot. I was delighted to learn this week that Yann LeCun has also joined. I’ve been consistently impressed by the thoughtful approach the Korean government has taken toward AI, with an emphasis on investment and innovation and a realistic understanding of risks without being distracted by science-fiction scenarios of harm.
I’ve advised many countries to build AI for the sectors where they’re strong. For example, I wrote previously that by investing in sectors like tourism and certain industrial areas, Thailand can do projects more efficiently than I can in Silicon Valley. South Korea’s tech ecosystem gives it a foundation to move fast even across numerous sectors. This emphasizes the long-term value for countries to become good at tech, because tech is now pervasive and affects all industries.
Korea has a very strong local software ecosystem. For example, the dominant search engine is not Google or Bing, but Naver (a Korean company). The dominant messaging system is not WhatsApp or WeChat, but KakaoTalk. With local tech giants Naver and Kakao offering email, mobile payment, cloud computing, ride sharing, and other services, the country has many sophisticated tech businesses. Additionally, SK hynix and Samsung are advanced semiconductor manufacturers. It also has a thriving entrepreneurship ecosystem, including Upstage, a language modeling startup, which taught a course with us on “Pretraining LLMs.” Finally, the Korean institutions Seoul National University, which I visited last year, and KAIST have global reputations.
Korea has a highly educated population, highly skilled software engineers, and a thriving set of software products. This gives it a fantastic foundation to embrace the next generation of AI. After meeting with businesses in retail, construction, insurance, cosmetics, telecoms, and other industries, I was delighted by the wide variety of opportunities many companies are pursuing across different industry sectors.
Lastly, Korea is known globally for its K-pop. Meeting Bang Si-Hyuk, the chairman of HYBE, which manages the superstar singing group BTS, and learning how the company operates was a real treat!
That’s why I’ve traveled to South Korea four times since last year. My venture studio AI Fund, which collaborates with many Korean companies, has benefited tremendously from the advice of many South Koreans, including Taizo Son, Changmook Kang, Hyungjun Kim, Sung Kim, JP Lee, Ian Park, and Alice Oh. I look forward to doing more in, and with, South Korea!
[Original text (with links): www.deeplearning.ai/the-batch/issue-265/ ] -
Andrew Ng
@AndrewYNg
We just released the final two courses of AI Python for Beginners! The complete set of four courses is now available and remains free for a limited time.
They teach how to write code (a) Using AI-assistance, which is where the field is going, and (b) to take advantage of generative AI, which allows you to quickly do valuable things with code.
If you're considering learning to code, AI has made this a great time to jump in. Or if you know someone who is considering learning, please recommend these courses!
www.deeplearning.ai/short-courses/ai-python-for-beginners/ -
Andrew Ng
@AndrewYNg
There’s still time to stop California’s SB 1047 from becoming law. For @TIME, I wrote about why this bill would hinder developers and actually make AI less safe. We should be regulating harmful applications of AI, not general-purpose AI models. time.com/collection/time100-voices/7016134/california-sb-…
-
Andrew Ng
@AndrewYNg
After a recent price reduction by OpenAI, GPT-4o tokens now cost $4 per million tokens (using a blended rate that assumes 80% input and 20% output tokens). GPT-4 cost $36 per million tokens at its initial release in March 2023. This price reduction over 17 months corresponds to about a 79% drop in price per year. (4/36 = (1 - p)^{17/12})
As you can see, token prices are falling rapidly! One force that’s driving prices down is the release of open weights models such as Llama 3.1. If API providers, including startups Anyscale, Fireworks, Together AI, and some large cloud companies, do not have to worry about recouping the cost of developing a model, they can compete directly on price and a few other factors such as speed.
Further, hardware innovations by companies such as Groq (a leading player in fast token generation), Samba Nova (which serves Llama 3.1 405B tokens at an impressive 114 tokens per second), and wafer-scale computation startup Cerebras (which just announced a new offering this week), as well as the semiconductor giants NVIDIA, AMD, Intel, and Qualcomm, will drive further price cuts.
When building applications, I find it useful to design to where the technology is going rather than only where it has been. Based on the technology roadmaps of multiple software and hardware companies — which include improved semiconductors, smaller models, and algorithmic innovation in inference architectures — I’m confident that token prices will continue to fall rapidly.
This means that even if you build an agentic workload that isn’t entirely economical, falling token prices might make it economical at some point. As I wrote previously, being able to process many tokens is particularly important for agentic workloads, which must call a model many times before generating a result. Further, even agentic workloads are already quite affordable for many applications. Let's say you build an application to assist a human worker, and it uses 100 tokens per second continuously: At $4/million tokens, you'd be spending only $1.44/hour – which is significantly lower than the minimum wage in the U.S. and many other countries.
So how can AI companies prepare?
- First, I continue to hear from teams that are surprised to find out how cheap LLM usage is when they actually work through cost calculations. For many applications, it isn’t worth too much effort to optimize the cost. So first and foremost, I advise teams to focus on building a useful application rather than on optimizing LLM costs.
- Second, even if an application is marginally too expensive to run today, it may be worth deploying in anticipation of lower prices.
- Finally, as new models get released, it might be worthwhile to periodically examine an application to decide whether to switch to a new model either from the same provider (such as switching from GPT-4 to the latest GPT-4o-2024-08-06) or a different provider, to take advantage of falling prices and/or increased capabilities.
Because multiple providers now host Llama 3.1 and other open-weight models, if you use one of these models, it might be possible to switch between providers without too much testing (though implementation details — specifically quantization, does mean that different offerings of the model do differ in performance). When switching between models, unfortunately, a major barrier is still the difficulty of implementing evals, so carrying out regression testing to make sure your application will still perform after you swap in a new model can be challenging. However, as the science of carrying out evals improves, I’m optimistic that this will become easier.
[Original text (with links): www.deeplearning.ai/the-batch/issue-264/ ] -
Andrew Ng
@AndrewYNg
Explore state-of-the-art multimodal prompting in our new short course Large Multimodal Model Prompting with Gemini, taught by Erwin Huizenga in collaboration with @googlecloud.
One interesting insight from this course: with multimodal models, prompt structure matters significantly. Placing text inputs, such as a patient's medical history, before image inputs, like an X-ray, can enhance the model's ability to contextualize and interpret visual data effectively. In other contexts, such as image captioning, you may get better results by putting the image first. Multimodal models behave differently than text-only LLMs, and effective prompting for models varies depending on the model you’re using. In this course you’ll learn how to effectively prompt Gemini models.
Gemini's multimodal capabilities also enable new approaches in AI application development, for example:
- The Gemini library handles various video formats (MP4, MOV, MPEG), streamlining applications using these formats.
- Large context window (up to 1 million tokens) enables processing of extensive content, like analyzing multiple 50-minute videos simultaneously.
- Function calling feature integrates real-time data (e.g., current exchange rates) into model responses.
The course demonstrates building multimodal applications with real-world examples including document analyzers that reason across text and graphs simultaneously, video content extractors that find and timestamp specific information from multiple hours of footage, and automated expense report systems processing receipt images while cross-referencing company policies.
Sign up here: www.deeplearning.ai/short-courses/large-multimodal-model-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Really fun hackathon, and it was great to see 30 creative Agentic AI projects, all built in a day. Well done to all the hackathon participants, and thank you @HenryYin_ , @AlexReibman and @agihouse_org for having me!
·
AGI House
@agihouse_org
The legendary Andrew Ng @AndrewYNg keynoting at our agent hackathon
-
Andrew Ng
@AndrewYNg
I've been playing with @SambaNovaAI's API serving fast Llama 3.1 405B tokens. Really cool to see leading model running at speed. Congrats to Samba Nova for hitting a 114 tokens/sec speed record (and also thanks @KunleOlukotun for getting me an API key!) sambanova.ai/blog/speed-record-on-llama-3.1-405b
-
Andrew Ng
@AndrewYNg
I am thrilled to announce that Dan Maloney is becoming the Chief Executive Officer (CEO) of LandingAI, and I will step into the company’s Executive Chairman role, where I will continue to focus on Visual AI deep tech with our team, including agentic vision.
Since Dan joined two years ago as the company’s Chief Operating Officer (COO), his leadership across the company has been instrumental to LandingAI serving more customers and better than ever. Dan played an integral role in all of our key initiatives, including the launch of our self-service offering with LandingLens available for any vision application builder to use, establishing dozens of strategic and channel partnerships, increasing our focus on serving developers, and improving our operational excellence; even as we continue to innovate on Visual AI technology.
I look forward to continuing LandingAI’s journey with Dan, with our fantastic team, and with all our wonderful partners and users!
Text of full announcement below. landing.ai/blog/charting-new-horizons-landingai-welcomes-… -
Andrew Ng
@AndrewYNg
I’m encouraged at the progress of the U.S. government at moving to stem harmful AI applications. Two examples are the new Federal Trade Commission (FTC) ban on fake product reviews and the DEFIANCE Act, which imposes punishments for creating and disseminating non-consensual deepfake porn. Both rules take a sensible approach to regulating AI insofar as they target harmful applications rather than general-purpose AI technology.
The best way to ensure AI safety is to regulate it at the application level rather than the technology level. This is important because the technology is general-purpose and its builders (such as a developer who releases an open-weights foundation model) cannot control how someone else might use it. If, however, someone applies AI in a nefarious way, we should stop that application.
Even before generative AI, fake reviews were a problem on many websites, and many tech companies dedicate considerable resources to combating them. A telltale sign of old-school fake reviews is the use of similar wording in different reviews. AI’s ability to automatically paraphrase or rewrite is making fake reviews harder to detect.
Importantly, the FTC is not going after the makers of foundation models for fake reviews. The provider of an open weights AI model, after all, can’t control what someone else uses it for. Even if one were to try to train a model to put up guardrails against writing reviews, I don’t know how it could distinguish between a real user of a product asking for help writing a legitimate review and a spammer who wanted a fake review. The FTC appropriately aims to ban the application of fake reviews along with other deceptive practices such as buying positive reviews.
The DEFIANCE Act, which passed unanimously in the Senate (and still requires passage in the House of Representatives before the President can sign it into law) imposes civil penalties for the creating and distributing non-consensual, deepfake porn. This disgusting application is harming many people including underage girls. While many image generation models do have guardrails against generating porn, these guardrails often can be circumvented via jailbreak prompts or fine-tuning (for models with open weights).
Again, DEFIANCE regulates an application, not the underlying technology. It aims to punish people who engage in the application of creating and distributing non-consensual intimate images, regardless of how they are generated — whether the perpetrator uses a diffusion model, a generative adversarial network, or Microsoft Paint to create an image pixel by pixel.
I hope DEFIANCE passes in the House and gets signed into law. Both rules guard against harmful AI applications without stifling AI technology itself (unlike California’s poorly designed SB-1047), and they offer a good model for how the U.S. and other nations can protect citizens against other potentially harmful applications.
[Original text (with links): www.deeplearning.ai/the-batch/issue-263/ ] -
Andrew Ng
@AndrewYNg
Build and customize complex AI applications with a flexible framework in this new short course, Building AI Applications with Haystack. Created in collaboration with @deepset_ai, and taught by @tuanacelik, who is the developer relations lead for Haystack at deepset.
Generative AI technology is changing rapidly and it can be challenging to integrate APIs from different LLMs, vector databases, and various tools such as web search. In this course, you will learn how to use the Haystack framework to make your development process more modular, allowing you to manage complexity and focus more on building your application.
In detail, you’ll:
- Build a RAG pipeline using Haystack’s main building blocks – components, pipelines, and document stores.
- Create custom components in your pipeline by building a Hacker News summarizer that extends your app’s ability to access APIs.
- Use conditional routing to create a branching pipeline with a fallback to web search mechanism when the LLM does not have the necessary context to respond to the user's query.
- Build a self-reflecting agent for named entity recognition that loops using an output validator custom component.
- Create a chat agent using OpenAI's function-calling capabilities which allow you to provide Haystack pipelines as tools to the LLM, enhancing that agent's capabilities.
By the end of this course, you will learn a high-level orchestration framework that can help make your applications flexible, extendible, and maintainable, even as the technology stack changes, new user needs arise, and you add new features to your application.
Please sign up here: www.deeplearning.ai/short-courses/building-ai-application…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Thank you @RepZoeLofgren for speaking out against the anti-open source, anti-innovation bill SB-1047.
She is the ranking member of the House Science, Space and Technology Committee. Her staff concludes: "the problematic core concerns remain: there is little evidentiary basis for the bill; the bill would negatively affect open-source development by applying liability to downstream use; it uses arbitrary thresholds not backed in science." [my boldface]
Lets all keep fighting to protect open source AI.
lofgren.house.gov/sites/evo-subsites/lofgren.house.gov/fi… -
Andrew Ng
@AndrewYNg
When entrepreneurs build a startup, it is often their speed and momentum that gives them a shot at competing with the tech behemoths. This is true of countries as well.
I was recently in Thailand, where I was delighted to see tremendous momentum building in AI (and sip the best Thai ice tea I’ve ever tasted). Even though Thailand is not as advanced in AI technology or applications as leading tech countries, the enthusiasm for building AI throughout government, corporations, and academia was thrilling. I came away heartened that AI’s benefits will be spread among many countries and convinced that one's level of AI development right now matters less than your momentum toward increasing it.
Seeing the momentum behind AI in Thailand — where the per capita GDP is around one fifth that of Japan, and one tenth that of the United States — left me feeling that any country, company, or person has a shot at doing meaningful work in the field. While advanced economies such as the U.S. and China are still in the lead, generative AI has made the playing field more level. Foundation models, especially those with open weights, are significantly lowering the barriers to building meaningful AI projects. In Thailand, a lot of people I met weren’t just talking about AI, they were rolling up their sleeves and building. That buys a nation a lot more momentum than just talk.
I met with Prime Minister Srettha Thavisin and his Ministers of Higher Education and Education (primary/secondary) along with many staffers. It was delightful to hear the PM speak of his enthusiasm for AI. The ministers discussed how to (i) provide AI training and (ii) use AI to improve education in a variety of subjects. Happily, the focus was on creating value while thinking through realistic risks like AI’s potential to proliferate misinformation, and not a single person asked me about whether AI will lead to human extinction!
I also met with many business leaders and enjoyed seeing a rapid pace of experimentation with AI. KBTG, an affiliate of the country’s leading digital bank KBank, is working on a financial chatbot advisor, AI-based identity verification for anti-fraud, AI for auto insurance, and a Thai-language financial large language model. These features are growing mobile banking and increasing financial access. Many business leaders in other sectors, too, have asked their teams to run experiments. There are many AI applications yet to be built in industrial sectors, tourism, trade, and more! (KBTG is an investor in AI Fund, which I lead.)
I often visit universities in both developed and developing economies, and I’ve been surprised to see that universities in developing economies sometimes adopt AI faster. At Chulalongkorn University (known as Chula), I met with the University President Bundhit Eua-arporn and Director of Chula AI Professor Proadpran Punyabukkana. Chula AI has rolled out campus-wide training in generative AI for faculty, staff, and students. In addition, it supports building AI applications such as AI screening for depression and gastrointestinal cancer.
It takes years to build up advanced technology. But momentum matters, and there will be many rewards along the journey. There’s no time like the present to start building!
[Original text: www.deeplearning.ai/the-batch/issue-262/ ] -
Andrew Ng
@AndrewYNg
Learn a development pattern to systematically improve the accuracy and reliability of LLM applications in our new short course, Improving Accuracy of LLM Applications, built in partnership with @LaminiAI and @Meta, and taught by Lamini’s CEO @realSharonZhou, and Meta’s Senior Director of Partner Engineering, @asangani7. (Disclosure: I am an investor in Lamini.)
The path to tuning an LLM application can be complex. In this course, you'll learn a systematic sequence of steps for improving accuracy by reducing hallucinations:
- Create an evaluation dataset to measure model accuracy
- Add prompt engineering and self-reflection
- Fine-tune your model including "memory-tuning" which is a new method of embedding facts in an LLM
Using the Llama 3-8B parameter model, you will:
- Build a text-to-SQL agent with a custom schema and simulate situations where it hallucinates
- Understand the difference between instruction fine-tuning, which gives pre-trained LLMs instructions to follow, and memory fine-tuning
- See how Performance-Efficient Fine-tuning (PEFT) techniques like Low-Rank Adaptation (LoRA) reduce training time by 100x and Mixture of Memory Experts (MoME) reduces it even further
I appreciate Meta releasing the Llama's family of open models -- this course gives an example of the unique type of work that developers can do with such models.
Please sign up here: www.deeplearning.ai/short-courses/improving-accuracy-of-l…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
It is very rare for the U.S. Federal government to chime in on state-level legislation. I'm glad that Congressman @RoKhanna is speaking out about why California's proposed SB 1047 is a bad idea.
He writes that he is "concerned that the bill as currently written would be ineffective, punishing of individual entrepreneurs and small businesses, and hurt California’s spirit of innovation." I agree with him.
khanna.house.gov/media/press-releases/khanna-statement-sb… -
Andrew Ng
@AndrewYNg
I’m speaking on a panel this Thursday (3pm PT) about Stratospheric Aerosol Injection (SAI). SAI is a potential approach to reducing global warming, and AI climate modeling has an important role to play in understanding its impact. I look forward to discussing the atmospheric science and social/political factors of SAI with @chrfield @DKeithClimate @DougMacMartin and Simone Tilmes.
Please register here to join us! stanford.zoom.us/webinar/register/WN_iR7jyDR3Qo6PNd0y6Azbdg -
Andrew Ng
@AndrewYNg
There is an chorus of voices across academia, business and even government concerned about California's proposed anti-open source, anti-innovation bill SB 1047.
@martin_casado has a nice thread summarizing some of these. Thank you @russellwald, @vishalmisra, Ion Stoica, many people from the University of California community, @AnimaAnandkumar, @drfeifei, @garrytan, @AnjneyMidha, Zoe Lofgren and many many others for speaking out to explain why this bill will hurt AI innovation without actually increasing safety.
If anything, by hurting open source and thus hampering researchers' ability to study cutting-edge models to identify problems, I believe SB1047 is more likely to make AI less safe.
martin_casado
@martin_casado
.@Scott_Wiener continues to falsely claim narrow opposition to SB 1047. When in reality there is massive public outcry across research, academic, public and private business and finance. Here is a mega roundup of recent announcements that fully debunk the Senator's claims 🧵
-
Andrew Ng
@AndrewYNg
I'm teaching a new course! AI Python for Beginners is a series of four short courses that teach anyone to code, regardless of current technical skill. We are offering these courses free for a limited time.
Generative AI is transforming coding. This course teaches coding in a way that’s aligned with where the field is going, rather than where it has been:
(1) AI as a Coding Companion. Experienced coders are using AI to help write snippets of code, debug code, and the like. We embrace this approach and describe best-practices for coding with a chatbot. Throughout the course, you'll have access to an AI chatbot that will be your own coding companion that can assist you every step of the way as you code.
(2) Learning by Building AI Applications. You'll write code that interacts with large language models to quickly create fun applications to customize poems, write recipes, and manage a to-do list. This hands-on approach helps you see how writing code that calls on powerful AI models will make you more effective in your work and personal projects.
With this approach, beginning programmers can learn to do useful things with code far faster than they could have even a year ago.
Knowing a little bit of coding is increasingly helping people in job roles other than software engineers. For example, I've seen a marketing professional write code to download web pages and use generative AI to derive insights; a reporter write code to flag important stories; and an investor automate the initial drafts of contracts.
With this course you’ll be equipped to automate repetitive tasks, analyze data more efficiently, and leverage AI to enhance your productivity.
If you are already an experienced developer, please help me spread the word and encourage your non-developer friends to learn a little bit of coding.
I hope you'll check out the first two short courses here! www.deeplearning.ai/short-courses/ai-python-for-beginners/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I wrote last week about why working on a concrete startup or project idea — meaning a specific product envisioned in enough detail that we can build it for a specific target user — lets you go faster. In this letter, I’d like to share some best practices for identifying promising ideas.
AI Fund, which I lead, works with many corporate partners to identify ideas, often involving applications of AI to the company’s domain. Because AI is applicable to numerous sectors such as retail, energy, logistics and finance, I’ve found working with domain experts who know these areas well immensely helpful for identifying what applications are worth building in these areas.
Our brainstorming process starts with recommending that a large number of key contributors at our partner corporation (at least 10 but sometimes well over 100) gain a non-technical, business-level understanding of AI and what it can and can’t do. Taking deeplearning.ai’s “Generative AI for Everyone” course is a popular option, after which a company is well positioned to assign a small team to coordinate a brainstorming process, followed by a prioritization exercise to pick what to work on. The brainstorming process can be supported by a task-based analysis of jobs in which we decompose employees’ jobs into tasks to identify which ones might be automated or augmented using AI.
Here are some best practices for these activities:
(i) Trust the domain expert’s gut. A domain expert who has worked for years in a particular sector will have well honed instincts that let them make leaps that would take a non-expert weeks of research.
Let’s say we’re working with a financial services expert and have developed a vague idea (“build a chatbot for financial advice”). To turn this into a concrete idea, we might need to answer questions such as what areas of finance to target (should we focus on budgeting, investing, or insurance?) and what types of user to serve (fresh graduates, mortgage applicants, new parents, or retirees?) Even a domain expert who has spent years giving financial advice might not know the best answer, but a choice made via their gut gives a quick way to get to one plausible concrete idea. Of course, if market-research data can be obtained quickly to support this decision, we should take advantage of it. But to avoid slowing down too much, we’ve found that experts’ gut reactions work well and are a quick way to make decisions.
So, if I’m handed a non-concrete idea, I often ask a domain expert to use their gut — and nothing else — to quickly make decisions as needed to make the idea concrete. The resulting idea is only a starting point to be tweaked over time. If, in the discussion, the domain expert picks one option but seems very hesitant to disregard a different option, then we can also keep the second option as a back-up that we can quickly pivot to if the initial one no longer looks promising.
(ii) Generate many ideas. I usually suggest coming up with at least 10 ideas; some will come up with over 100, which is even better. The usual brainstorming advice to go for volume rather than quality applies here. Having many ideas is particularly important when it comes to prioritization. If only one idea is seriously considered — sometimes this happens if a senior executive has an idea they really like and puts this forward as the “main” idea to be worked on — there’s a lot of pressure to make this idea work. Even if further investigation discovers problems with it — for example, market demand turns out to be weak or the technology is very expensive to build — the team will want to keep trying to make it work so we don’t end up with nothing.
In contrast, when a company has many ideas to choose from, if one starts to look less interesting, it’s easy to shift attention to a different one. When many ideas are considered, it’s easier to compare them to pick the superior ones. As explained in the book Ideaflow, teams that generate more ideas for evaluation and prioritization end up with better solutions.
Because of this, I’ve found it helpful to run a broad brainstorming process that involves many employees. Specifically, large companies have many people who collectively have a lot of wisdom regarding the business. Having a small core team coordinate the gathering of ideas from a large number of people lets us tap into this collective fountain of invention. Many times I’ve seen a broad effort (involving, say, ~100 people who are knowledgeable about the domain and have a basic understanding of AI) end up with better ideas than a narrow one (involving, say, a handful of top executives).
(iii) Make the evaluation criteria explicit. When evaluating and prioritizing, clear criteria for scoring and ranking ideas helps the team to judge ideas more consistently. Business value and technical feasibility are almost always included. Additionally, many companies will prioritize projects that can be a quick win (to build momentum for their overall AI efforts) or support certain strategic priorities such as growth in a particular part of the business. Making such criteria explicit can help during the idea-generation phase, and it’s critical when you evaluate and prioritize.
In large companies, it can take a few weeks to go through a process to gather and prioritize ideas, but this pays off well in identifying valuable, concrete ideas to pursue. AI isn’t useful unless we find appropriate ways to apply it, and I hope these best practices will help you to generate great AI application ideas to work on.
[Original text: www.deeplearning.ai/the-batch/issue-260/ ] -
Andrew Ng
@AndrewYNg
Learn how embedding models are built, trained, and used in semantic search systems in this new short course, Embedding Models: From Architecture to Implementation, created with @vectara and taught by @ofermend.
Many LLM apps use a single embedding model for both questions and answers. This can lead to issues such as retrieving results that are similar to the question itself rather than relevant answers. With a dual encoder architecture, you can use separate embedding models for questions and answers to better match questions with appropriate answers.
In this course, you will use, build, and train a dual encoder model.
You’ll also learn:
- What are word embeddings, and how they are used
- The evolution of embeddings to BERT, where embeddings take into account each word's surrounding context
- How a contrastive loss is used to train a dual encoder model with one encoder trained to embed questions and the other responses.
- How to analyze a dual encoder’s effect on search relevance and compare it to a retrieval process with a single encoder.
Please sign up here! www.deeplearning.ai/short-courses/embedding-models-from-a…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
AI’s usefulness in a wide variety of applications creates many opportunities for entrepreneurship. Here, I’d like to share what might be a counter-intuitive best practice that I’ve learned from leading AI Fund, a venture studio that has built dozens of startups with extraordinary entrepreneurs. When it comes to building AI applications, we strongly prefer to work on a concrete idea, meaning a specific product envisioned in enough detail that we can build it for a specific target user.
Some design philosophies say you shouldn’t envision a specific product from the start. Instead, they recommend starting with a problem to be solved and then carefully studying the market before you devise a concrete solution. There’s a reason for this: The more concrete or precise your product specification, the more likely it is to be off-target. However, I find that having something specific to execute toward lets you go much faster and discover and fix problems more rapidly along the way. If the idea turns out to be flawed, rapid execution will let you discover the flaws sooner, and this knowledge and experience will help you switch to a different concrete idea.
One test of concreteness is whether you’ve specified the idea in enough detail that a product/engineering team could build an initial prototype. For example, “AI for livestock farming” is not concrete; it’s vague. If you were to ask an engineer to build this, they would have a hard time knowing what to build. Similarly, “AI for livestock tracking in farming” is still vague. There are so many approaches to this that most reasonable engineers wouldn’t know what to build. But “Apply face recognition to cows so as to recognize individual cows and monitor their movement on a farm” is specific enough that a good engineer could quickly choose from the available options (for example, what algorithm to try first, what camera resolution to use, and so on) to let us relatively efficiently assess:
- Technical feasibility: For example, do face recognition algorithms developed for human faces work for cows? (It turns out that they do!)
- Business feasibility: Does the idea add enough value to be worth building? (Talking to farmers might quickly reveal that solutions like RFID are easier and cheaper.)
Articulating a concrete idea — which is more likely than a vague idea to be wrong — takes more courage. The more specific an idea, the more likely it is to be a bit off, especially in the details. The general area of AI for livestock farming seems promising, and surely there will be good ways to apply AI for livestock. In contrast, specifying a concrete idea, which is much easier to invalidate, is scary.
The benefit is that the clarity of a specific product vision lets a team execute much faster. One strong predictor of how likely a startup is to succeed is the speed with which it can get stuff done. This is why founders with clarity of vision tend to be desired; clarity helps drive a team in a specific direction. Of course, the vision has to be a good one, and there’s always a risk of efficiently building something that no one wants to buy! But a startup is unlikely to succeed if it meanders for too long without forming a clear, concrete vision.
Building toward something concrete — if you can do so in a responsible way that doesn’t harm others — lets you get critical feedback more efficiently and, if necessary, switch directions sooner. (An earlier letter in The Batch also discussed when it’s better to go with a “Ready, Fire, Aim” approach to projects.) One factor that favors this approach is the low cost of experimenting and iterating. This is increasingly the case for many AI applications, but perhaps less so for deep-tech AI projects.
I realize that this advice runs counter to common practice in design thinking, which warns against leaping to a solution too quickly, and instead advocates spending time understanding end-users, deeply understanding their problems, and brainstorming a wide range of solutions. If you’re starting without any ideas, then such an extended process can be a good way to develop good ideas. Further, keeping ideas open-ended can be good for curiosity-driven research, where investing to pursue deep tech with only a vague direction in mind can pay huge dividends over the long term.
If you are thinking about starting a new AI project, consider whether you can come up with a concrete vision to execute toward. Even if the initial vision turns out not to be quite right, rapid iteration will let you discover this sooner, and the learnings will let you switch to a different concrete idea.
[Original text: www.deeplearning.ai/the-batch/issue-259/ ] -
Andrew Ng reposted
Garry Tan
@garrytan
"This is one of those moments where such laws, not just in California but around the world, could have a huge impact on whether entrepreneurs and startups are allowed to keep on innovating." —@AndrewYNg
-
Andrew Ng
@AndrewYNg
Learn to train an LLM with distributed data while ensuring privacy using federated learning in a new two-part short course, Intro to Federated Learning and Federated Fine-tuning of LLMs with Private Data, created with @flwrlabs and taught by @daniel_janes and @niclane7.
Federated learning allows a single model to be trained across multiple devices, such as phones, or multiple organizations, such as hospitals, without the need to share data to a central server.
This two-part course gives you an introduction to federated learning, and then teaches you how to fine-tune your large language model with distributed data using Flower Lab’s open source federated learning framework.
You’ll learn:
- How to use federated learning to train a variety of models, ranging from speech and vision models to LLMs, across distributed data while offering data privacy options to users and organizations.
- Privacy Enhancing Technologies like differential privacy (DP), which obscures individual data by adding calibrated noise to query results.
- Two variants of differential privacy - Central and Local - and how to choose depending on your use case.
- How to measure and decrease bandwidth usage to make federated learning more practical and efficient with techniques like using pre-trained models and Parameter-Efficient Fine-Tuning
- How federated LLM fine-tuning reduces the risk of leaking training data.
Sign up here! www.deeplearning.ai/short-courses/federated-learning/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Thank you Meta and the Llama team for your huge contributions to open-source! Llama 3.1 with increased context length and improved capabilities is a wonderful gift to everyone.
I hope foolish regulations don't like California's proposed SB1047 don't stop such innovations. -
Andrew Ng
@AndrewYNg
New short course on Pretraining LLMs! Developed with @UpstageAI and taught by their CEO @hunkims and CSO @echojuliett.
While prompting or fine-tuning existing models works well for many general language tasks, pretraining is valuable for specialized domains or languages with limited representation in current models.
This course walks you through the LLM pretraining pipeline:
1. Data preparation: Learn to source, clean, and prepare training data using HuggingFace.
2. Model architecture: Configure transformer networks, including modifying existing models.
3. Training: Set up and run training using open-source libraries.
4. Evaluation: Benchmark performance using popular evaluation strategies.
As an example use case, you'll also compare the output of a base model with its fine-tuned and further pretrained variants, to see the impact of pretraining on a model's ability to write Python.
The course also explores an innovative technique called depth up-scaling, which Upstage used to train their Solar model family, reducing pretraining compute costs by up to 70%. This technique works by first duplicating layers of a smaller pretrained model to form a larger model, and then further pretraining the result.
Sign up here! www.deeplearning.ai/short-courses/pretraining-llms/This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Yann LeCun
@ylecun
Excellent argument by @AndrewYNg against the ignominious California regulation SB1047, which would essentially kill open source AI and significantly slow down or stop AI innovation.
Andrew Ng
@AndrewYNg
I continue to be alarmed at the progress of proposed California regulation SB 1047 and the attack it represents on open source and more broadly on AI innovation. As I wrote previously, this proposed law makes a fundamental mistake of regulating AI technology instead of AI applications, and thus would fail to make AI meaningfully safer. I’d like to explain why the specific mechanisms of SB 1047 are so pernicious to open source.
To be clear, there are routes that regulators should pursue to improve safety. For example, I would welcome outlawing nonconsensual deepfake pornography, standardizing watermarking and fingerprinting to identify generated content, and investing more in red teaming and other safety research. Unfortunately, the proposed bill pursues a less beneficial and more harmful path.
SB 1047’s purported goal is to ensure safety of AI models. It puts in place complex reporting requirements for developers who fine-tune models or develop models that cost more than $100 million to train. It is a vague, ambiguous law that imposes significant penalties for violations, creating a huge gray zone in which developers can’t be sure how to avoid breaking the law. This will paralyze many teams.
You can read the latest draft of the law online. I’ve read through it carefully, and I find it ambiguous and very hard to follow.
Developers who try to navigate the law’s complex requirements face what feels like a huge personal risk. It requires that developers submit, under penalty of perjury, a certification of compliance with the requirements of the law. But when the requirements are complex, hard to understand, and can even shift according to the whims of an unelected body (more on this below), how do we ensure we are in compliance?
For example, the certification must include many different sections. One is an analysis of “the nature and magnitude of critical harms … the model might reasonably cause or enable.” But given that even leading AI researchers aren’t sure what harms models might cause or enable, how is a team of developers supposed to figure this out and declare — under penalty of perjury — that they meet this requirement?
Further, some developers will be required to implement “protections to prevent … misuse of, or unsafe post-training modifications of, the covered model and all covered model derivatives … that are appropriate in light of the risks associated with the covered model, including from advanced persistent threats or other sophisticated actors.” Even leading AI researchers don’t agree on how best to “protect” AI models against these supposed risks, or what would be “appropriate.” So how are developers supposed to figure out how to comply with this requirement?
This creates a scary situation for developers. Committing perjury could lead to fines and even jail time. Some developers will have to hire expensive lawyers or consultants to advise them on how to comply with these requirements. (I am not a lawyer and am not giving legal advice, but one way to try to avoid perjury is to show that you are relying on expert advice, to demonstrate that you had no intent to lie.) Others will simply refrain from releasing cutting-edge AI products.
If this law passes, the fear of a trial by a jury — leading to a verdict that can be very unpredictable and with significant penalties in the event of a conviction — will be very real. What if someone releases a model today after taking what they genuinely felt were reasonable safeguards, but a few years later, when views on AI technology might have shifted, some aggressive prosecutor manages to convince a jury that whatever they did was not, in hindsight, “reasonable”?
Reasonableness is ambiguous and its legal interpretation can depend on case law, jury instructions, and common facts, among other things. This makes it very hard to ensure that what a developer does today will be deemed reasonable by a future jury. (For more on this, see Context Fund’s analysis of SB 1047. [URLs in article linked to below.])
One highly placed lawyer in the California government who studied this law carefully told me they found it hard to understand. I invite you to read it and judge for yourself — if you find the requirements clear, you might have a brilliant future as a lawyer!
Adding to the ambiguity, the bill would create a Frontier Model Division (FMD) with a five-person board that has the power to dictate standards to developers. This small board would be a great target for lobbying and regulatory capture. (Bill Gurley has a great video on regulatory capture.) The unelected FMD can levy fees on developers to cover its costs. It can arbitrarily change the computation threshold at which fine-tuning a model becomes subject to its oversight. This can lead to even small teams being required to hire an auditor to check for compliance with an ambiguous safety standard.
These provisions don’t ensure that AI is safe. They create regulatory uncertainty, and more opportunities for vested interests wishing to stifle open-source to lobby for shifts in the requirements that raise the cost of compliance. This would lock out many teams that don’t have a revenue stream — specifically, many open-source contributors — that would let them pay for lobbyists, auditors, and lawyers to help ensure they comply with these ambiguous and unreasonable requirements.
Open source is a wonderful force that is bringing knowledge and tools to many people, and is a key pillar of AI innovation. I am dismayed at the concerted attacks on it. Make no mistake, there is a fight in California right now for the future health of open source. I am committed to doing what I can to preserve open source, but I don’t assume that the pro-open source side will prevail. I hope you will join me in speaking out against SB 1047 and other laws that threaten to stifle open source.
[Original text (with links): www.deeplearning.ai/the-batch/issue-257/ ] -
Andrew Ng
@AndrewYNg
I continue to be alarmed at the progress of proposed California regulation SB 1047 and the attack it represents on open source and more broadly on AI innovation. As I wrote previously, this proposed law makes a fundamental mistake of regulating AI technology instead of AI applications, and thus would fail to make AI meaningfully safer. I’d like to explain why the specific mechanisms of SB 1047 are so pernicious to open source.
To be clear, there are routes that regulators should pursue to improve safety. For example, I would welcome outlawing nonconsensual deepfake pornography, standardizing watermarking and fingerprinting to identify generated content, and investing more in red teaming and other safety research. Unfortunately, the proposed bill pursues a less beneficial and more harmful path.
SB 1047’s purported goal is to ensure safety of AI models. It puts in place complex reporting requirements for developers who fine-tune models or develop models that cost more than $100 million to train. It is a vague, ambiguous law that imposes significant penalties for violations, creating a huge gray zone in which developers can’t be sure how to avoid breaking the law. This will paralyze many teams.
You can read the latest draft of the law online. I’ve read through it carefully, and I find it ambiguous and very hard to follow.
Developers who try to navigate the law’s complex requirements face what feels like a huge personal risk. It requires that developers submit, under penalty of perjury, a certification of compliance with the requirements of the law. But when the requirements are complex, hard to understand, and can even shift according to the whims of an unelected body (more on this below), how do we ensure we are in compliance?
For example, the certification must include many different sections. One is an analysis of “the nature and magnitude of critical harms … the model might reasonably cause or enable.” But given that even leading AI researchers aren’t sure what harms models might cause or enable, how is a team of developers supposed to figure this out and declare — under penalty of perjury — that they meet this requirement?
Further, some developers will be required to implement “protections to prevent … misuse of, or unsafe post-training modifications of, the covered model and all covered model derivatives … that are appropriate in light of the risks associated with the covered model, including from advanced persistent threats or other sophisticated actors.” Even leading AI researchers don’t agree on how best to “protect” AI models against these supposed risks, or what would be “appropriate.” So how are developers supposed to figure out how to comply with this requirement?
This creates a scary situation for developers. Committing perjury could lead to fines and even jail time. Some developers will have to hire expensive lawyers or consultants to advise them on how to comply with these requirements. (I am not a lawyer and am not giving legal advice, but one way to try to avoid perjury is to show that you are relying on expert advice, to demonstrate that you had no intent to lie.) Others will simply refrain from releasing cutting-edge AI products.
If this law passes, the fear of a trial by a jury — leading to a verdict that can be very unpredictable and with significant penalties in the event of a conviction — will be very real. What if someone releases a model today after taking what they genuinely felt were reasonable safeguards, but a few years later, when views on AI technology might have shifted, some aggressive prosecutor manages to convince a jury that whatever they did was not, in hindsight, “reasonable”?
Reasonableness is ambiguous and its legal interpretation can depend on case law, jury instructions, and common facts, among other things. This makes it very hard to ensure that what a developer does today will be deemed reasonable by a future jury. (For more on this, see Context Fund’s analysis of SB 1047. [URLs in article linked to below.])
One highly placed lawyer in the California government who studied this law carefully told me they found it hard to understand. I invite you to read it and judge for yourself — if you find the requirements clear, you might have a brilliant future as a lawyer!
Adding to the ambiguity, the bill would create a Frontier Model Division (FMD) with a five-person board that has the power to dictate standards to developers. This small board would be a great target for lobbying and regulatory capture. (Bill Gurley has a great video on regulatory capture.) The unelected FMD can levy fees on developers to cover its costs. It can arbitrarily change the computation threshold at which fine-tuning a model becomes subject to its oversight. This can lead to even small teams being required to hire an auditor to check for compliance with an ambiguous safety standard.
These provisions don’t ensure that AI is safe. They create regulatory uncertainty, and more opportunities for vested interests wishing to stifle open-source to lobby for shifts in the requirements that raise the cost of compliance. This would lock out many teams that don’t have a revenue stream — specifically, many open-source contributors — that would let them pay for lobbyists, auditors, and lawyers to help ensure they comply with these ambiguous and unreasonable requirements.
Open source is a wonderful force that is bringing knowledge and tools to many people, and is a key pillar of AI innovation. I am dismayed at the concerted attacks on it. Make no mistake, there is a fight in California right now for the future health of open source. I am committed to doing what I can to preserve open source, but I don’t assume that the pro-open source side will prevail. I hope you will join me in speaking out against SB 1047 and other laws that threaten to stifle open source.
[Original text (with links): www.deeplearning.ai/the-batch/issue-257/ ] -
Andrew Ng
@AndrewYNg
Learn to optimize RAG for cost and performance in our new short course, Prompt Compression and Query Optimization, created with @MongoDB and taught by @richmondalake.
This course teaches you to combine traditional database capabilities with vector search using MongoDB for RAG. You'll learn these techniques:
- Vector search: For semantic matching of user queries
- Filtering using metadata: Pre- and post-filtering to narrow search results
- Projections: Selecting only necessary fields to minimize data returned
- Boosting: Reranking results to improve relevance
- Prompt compression: Using a small LLM to compress context, significantly reducing token count and processing costs
These methods address scaling, performance, and security challenges in large-scale RAG applications.
You can sign up here: www.deeplearning.ai/short-courses/prompt-compression-and-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
As we reach the milestone of the 256th issue of The Batch, I’m reflecting on how AI has changed over the years and how society continues to change with it. As AI becomes more widely available, it’s clear that many people — developers and non-developers — will benefit from high-quality training to keep up with the changes and gain useful AI skills.
In my years of working in education, I’ve felt that the world has enough low-quality courses, newsletters, social media posts, and other forms of content. It’s possible to build a business churning out mediocre content in sufficient volume to attract a meaningful amount of attention, but I have no interest in doing that.
At deeplearning.ai, our core philosophy is to put learners first. Our team obsesses about how to create quality training or other programs that benefit people who want to learn about AI. We have intense debates about what tools to teach, which examples to include, even which partners to work with, based on what we think is best for learners.
For example, I recall vividly how, when working on the Machine Learning Specialization, our team spent ages debating whether to use row or column matrices. Both sides showed up with deep analysis of the pros and cons, made Powerpoint presentations to argue their case, and we spent hours debating over what was better for learners in terms of both ease of picking up the concepts as well as subsequently being able to use these skills with third-party machine learning libraries.
We don’t release a course unless we think it’s a good use of a learner’s time and we’d be proud to recommend it to our own friends and family members. Quality, of course, can mean a lot of things. I expect what we do to be technically accurate, useful, up to date, clear, and time-efficient for learners. And, if possible, fun!
We don’t always get it right, but we scrutinize learner feedback (one of my most important weekly routines is to study a dashboard that summarizes learner ratings of our courses) and work to make sure our courses serve learners well. And yes, we have a large-language model powered application that reads learner reviews to flag important issues quickly.
Earlier this year, we realized that some of the paid content we had launched was below our quality standard, and that I wouldn’t in good conscience recommend it to my friends or family members. Despite this content being profitable, we did what we felt was the right thing for learners. So we decided to retire that content and forgo the revenues, but we feel much better now for having done the right thing for learners.
When we teach courses with partners, we tell them our priorities are “learners first, partners second, ourselves last.” I’m grateful to the many wonderful companies and individuals that work with us to teach cutting-edge techniques, and given an opportunity we try to support our partners’ goals as well. But we never prioritize the interest of our educational partners over that of learners. Fortunately, our partners are onboard with this as well. We have a common goal to serve learners. Without their help, it would be difficult to teach many of the topics we do with high-quality content.
Quite a few companies have tried to offer to pay us to teach a course with them, but we’ve always said no. We work only with the companies that we think help us serve learners best, and are not interested in being paid to teach lower quality courses.
One reason I obsess about building quality training materials is that I think learning must be a habit. Learning a little every week is important to get through the volume of learning we all need, and additionally to keep up with changing technology. High-quality training that’s also fun supports a healthy learning habit!
Fun fact: In addition to taking online courses, I also read a lot. Recently I noticed that my digital reading app says I’ve been on a reading streak for 170 weeks. I’ve used the app for many years, but apparently I had broken and restarted my streak 170 weeks ago. What happened then? That was the week that my son was born, Coursera became a public company, and my grandfather died. While my life has had disruptions since then, I was happy to find that it takes a disruption of this magnitude to make me pause my learning habit for a week.
[Original text: www.deeplearning.ai/the-batch/issue-256/ ] -
Andrew Ng
@AndrewYNg
Shoutout to the team that built artificialanalysis.ai . Really neat site that benchmarks the speed of different LLM API providers to help developers pick which models to use. This nicely complements the LMSYS Chatbot Arena, Hugging Face open LLM leaderboards and Stanford's HELM that focus more on the quality of the outputs.
I hope benchmarks like this encourage more providers to work on fast token generation, which is critical for agentic workflows! -
Andrew Ng
@AndrewYNg
An often overlooked part of the AI supply chain is electricity to power our data centers.
@TheAESCorp is the leading provider of renewable energy to data centers, and is a also global leader in building technology to efficiently scale renewal energy projects. At AI Fund, we're thrilled to work with the visionary @AndresGluski and the AES team to co-build new AI companies that will help with the energy transition!
The AES Corporation
@TheAESCorp
We're excited to partner with @AI_Fund to co-build innovative companies that leverage AI to tackle today’s energy challenges.
With @AndrewYNg @AndresGluski and AI Fund we will continue to drive innovation forward that can help overcome bottlenecks in the energy transition. Learn more: aescorporation2023cr.q4web.com/press-releases/news-detail…
#aes #aifund #renewableenergy #sustainability #ai -
Andrew Ng
@AndrewYNg
As machine learning models grow in size, so too does their carbon footprint. As AI scales, it's important that we quantify and mitigate these emissions.
In our new short course Carbon Aware Computing for GenAI Developers you'll learn from @googlecloud Developer Advocate Nikita Namjoshi how to:
- Query the ElectricityMaps API for real-time data on regional electricity grid carbon intensity
- Route your model training jobs to data centers in regions primarily powered by low-carbon sources like wind, solar, hydro and nuclear
- Measure the carbon footprint of your ML training, inference, storage and API usage with the Google Cloud's Carbon Footprint tool
- Optimize training job scheduling to run when clean energy is most abundant in a given region.
You can sign up here: www.deeplearning.ai/short-courses/carbon-aware-computing-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
On Father’s Day last weekend, I sat with my daughter to help her practice solving arithmetic problems. To give her practice problems, I used OpenDevin, an open-source agentic coding framework, to write a Python script that generated questions that she enjoyed answering at her own pace. OpenDevin wrote the code much faster than I could have and genuinely improved my and my daughter’s day.
Six months ago, coding agents were a novelty. They still frequently fail to deliver, but I find that they’re now working well enough that they might be genuinely useful to more and more people!
Given a coding problem that’s specified in a prompt, the workflow for a coding agent typically goes something like this: Use a large language model (LLM) to analyze the problem and potentially break it into steps to write code for, generate the code, test it, and iteratively use any errors discovered to ask the coding agent to refine its answer. But within this broad framework, a huge design space and numerous innovations are available to experiment with. I’d like to highlight a few papers that I find notable:
- “AgentCoder: Multiagent-Code Generation with Iterative Testing and Optimisation,” Huang et al. (2024).
- “LDB: A Large Language Model Debugger via Verifying Runtime Execution Step by Step,” Zhong et al., (2024).
- “SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering,” Yang et al. (2024).
How can we test the code without requiring the user to write test cases? In a multi-agent system, each “agent” is an LLM prompted to play a particular role. An interesting result from AgentCoder shows that having separate agents for writing code and generating tests results in better performance than letting a single agent do both tasks. This is presumably because, if the agent writing the code is also responsible for writing the tests, the tests might be influenced by the code and fail to consider corner cases that the code does not cover.
When people think of testing code, many initially think of output testing, in which we see if the code generates the correct outputs to a specific set of test inputs. If the code fails a test, an LLM can be prompted to reflect on why the code failed and then to try to fix it. In addition to testing the output, the LDB method is helpful. LDB steps through the code and presents to the LLM values of the variables during intermediate steps of execution, to see if the LLM can spot exactly where the error is. This mimics how a human developer might step through the code to see where one of the computational steps went wrong, and so pinpoint and fix the problem.
A lot of agentic workflows mimic human workflows. Similar to other work in machine learning, if humans can do a task, then trying to mimic humans makes development much easier compared to inventing a new process. However, the authors of SWE-agent noticed that many tools that humans use for coding are very inefficient for agents. For example, giving an agent access to a bash shell and having it find a piece of code by executing numerous cd, ls, and cat commands is inefficient, even though humans can do this rapidly. Similarly, visual coding editors like VSCode, emacs, and vim are easy for humans to use, but hard for LLMs (or LMMs) to navigate. Because agents interact with computers differently than humans do, the authors found that building special-purpose tools (functions) to let an agent search, view, and edit codebases resulted in better performance.
One reason research into coding agents is making rapid progress is that their performance can be evaluated automatically and reliably. With benchmarks like HumanEval, MBPP, and SWE-bench, researchers can try out an idea and automatically test how often it generates correct code. In contrast, even though there’s considerable activity on AI research agents that search the web and synthesize an article (I’ve enjoyed using the open-source STORM system by Stanford's Yijia Shao et al.), they are hard to evaluate and this makes progress harder.
Github Copilot was released in 2021, and many developers have been getting coding help by prompting LLMs. The rapid evolution from that to more sophisticated coding agents is expanding how computers can help us with coding tasks, and the pace of progress is rapid. With these tools, I expect programming to become even more fun and more productive.
[Original text (with links): www.deeplearning.ai/the-batch/issue-254/ ] -
Andrew Ng
@AndrewYNg
Function calling is a powerful way to extend the capabilities of LLMs and AI agents by letting them use external tools. Our new short course Function calling and Data Extraction with LLMs, created with @NexusflowX and taught by @JiantaoJ and @VenkatKSrini, demonstrates how to prompt LLMs to form calls to external functions.
You'll work with NexusRavenV2-13B, a 13B parameter open-source model that excels in function calling tasks while still being small enough to host locally. Learn to use function calling to extract structured data from unstructured text and access web APIs, and build an end-to-end application that processes customer service transcripts. You'll learn how to build LLM-powered applications that can analyze feedback, automate data entry, and enhance search.
Please get started here: www.deeplearning.ai/short-courses/function-calling-and-da…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
One reason for machine learning’s success is that our field welcomes a wide range of work. I can’t think of even one example where someone developed what they called a machine learning algorithm and senior members of our community criticized it saying, “that’s not machine learning!” Indeed, linear regression using a least-squares cost function was used by mathematicians Legendre and Gauss in the early 1800s — long before the invention of computers — yet machine learning has embraced these algorithms, and we routinely call them “machine learning” in introductory courses!
In contrast, about 20 years ago, I saw statistics departments at a number of universities look at developments in machine learning and say, “that’s not really statistics.” This is one reason why machine learning grew much more in computer science than statistics departments. (Fortunately, since then, most statistics departments have become much more open to machine learning.)
This contrast came to mind a few months ago, as I thought about how to talk about agentic systems that use design patterns such as reflection, tool use, planning, and multi-agent collaboration to produce better results than zero-shot prompting. I had been involved in conversations about whether certain systems should count as “agents.” Rather than having to choose whether or not something is an agent in a binary way, I thought, it would be more useful to think of systems as being agent-like to different degrees. Unlike the noun “agent,” the adjective “agentic” allows us to contemplate such systems and include all of them in this growing movement.
More and more people are building systems that prompt a large language model multiple times using agent-like design patterns. But there’s a gray zone between what clearly is not an agent (prompting a model once) and what clearly is (say, an autonomous agent that, given high-level instructions, plans, uses tools, and carries out multiple, iterative steps of processing).
Rather than arguing over which work to include or exclude as being a true agent, we can acknowledge that there are different degrees to which systems can be agentic. Then we can more easily include everyone who wants to work on agentic systems. We can also encourage newcomers to start by building simple agentic workflows and iteratively make their systems more sophisticated.
In the past few weeks, I’ve noticed that, while technical people and non-technical people alike sometimes use the word “agent,” mainly only technical people use the word “agentic” (for now!). So when I see an article that talks about “agentic” workflows, I’m more likely to read it, since it’s less likely to be marketing fluff and more likely to have been written by someone who understands the technology.
Let’s keep working on agentic systems and keep welcoming anyone who wants to join our field!
[Original text: www.deeplearning.ai/the-batch/issue-253/ ] -
Andrew Ng
@AndrewYNg
New short course: Building Your Own Database Agent, created with Microsoft @Azure's @adriangs86. You'll learn to build an AI assistant that translates natural language questions into SQL queries. Querying with natural language empowers everyone in your organization, from business leaders to developers, to access data insights directly, using plain English.
You’ll use the Azure OpenAI Service and LangChain to implement retrieval augmented generation and function calling, and leverage the Assistants API. You'll gain hands-on experience building an agent that can reason over both CSV files and SQL databases.
Please sign up here! www.deeplearning.ai/short-courses/building-your-own-datab…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I think AI agentic machine translation has huge potential for improving over traditional neural machine translation, and am releasing as open-source a demonstration I'd been playing with as a fun weekend project.
Using an agentic workflow, this demonstration (i) Prompts an LLM to translate from one language to another, (ii) Reflects on the translation to come up with constructive suggestions, (iii) Uses the suggestions to refine the translation. In our limited testing, this is sometimes competitive with, and sometimes worse than, leading commercial providers.
But it gives a highly steerable translation system where by simply changing the prompt, you can specify the tone (formal/informal), regional variation (do you want Spanish as spoken in Spain or as spoken in Latin America?), and ensure consistent translation of terms (by providing a glossary).
This is not mature software. But I hope the open-source community can make agentic translation work much better. Given how a simple reflection workflow already gives decent results, I think there's significant headroom to make agentic translation much better.
Releasing an early software prototype like this is something new I decided to try to see if it is helpful to the developer community. I'd love any feedback on this.
Thanks to Joaquin Dominguez, @nedteneva, @JohnSanterre for help with this.
github.com/andrewyng/translation-agent -
Andrew Ng reposted
Yann LeCun
@ylecun
- Regulators should regulate applications, not technology.
- Regulating basic technology will put an end to innovation.
- Making technology developers liable for bad uses of products built from their technology will simply stop technology development.
- It will certainly stop the distribution of open source AI platforms, which will kill the entire AI ecosystem, not just startups, but also academic research.
- the strangest aspect of all this is that all of these regulations are based on completely hypothetical science fiction scenario that very, very few people in the field are plausible.
Andrew Ng
@AndrewYNg
The effort to protect innovation and open source continues. I believe we’re all better off if anyone can carry out basic AI research and share their innovations. Right now, I’m deeply concerned about California's proposed law SB-1047. It’s a long, complex bill with many parts that require safety assessments, shutdown capability for models, and so on.
There are many things wrong with this bill, but I’d like to focus here on just one: It defines an unreasonable “hazardous capability” designation that may make builders of large AI models liable if someone uses their models to do something that exceeds the bill’s definition of harm (such as causing $500 million in damage). That is practically impossible for any AI builder to ensure. If the bill is passed in its present form, it will stifle AI model builders, especially open source developers.
Some AI applications, for example in healthcare, are risky. But as I wrote previously, regulators should regulate applications rather than technology.
- Technology refers to tools that can be applied in many ways to solve various problems.
- Applications are specific implementations of technologies designed to meet particular customer needs.
For example, an electric motor is a technology. When we put it in a blender, an electric vehicle, dialysis machine, or guided bomb, it becomes an application. Imagine if we passed laws saying, if anyone uses a motor in a harmful way, the motor manufacturer is liable. Motor makers would either shut down or make motors so tiny as to be useless for most applications. If we pass such a law, sure, we might stop people from building guided bombs, but we’d also lose blenders, electric vehicles, and dialysis machines. In contrast, if we look at specific applications, like blenders, we can more rationally assess risks and figure out how to make sure they’re safe, and even ban classes of applications, like certain types of munitions.
Safety is a property of the application, not a property of the technology (or model), as @random_walker and @sayashk have pointed out. Whether a blender is a safe one can’t be determined by examining the electric motor. A similar argument holds for AI.
SB-1047 doesn’t account for this distinction. It ignores the reality that the number of beneficial uses of AI models is, like electric motors, vastly greater than the number of harmful ones. But, just as no one knows how to build a motor that can’t be used to cause harm, no one has figured out how to make sure an AI model can’t be adapted to harmful uses. In the case of open source models, there’s no known defense to fine-tuning to remove RLHF alignment. And jailbreaking work has shown that even closed-source, proprietary models that have been properly aligned can be attacked in ways that make them give harmful responses. Indeed, the sharp-witted @elder_plinius regularly tweets about jailbreaks for closed models. Kudos also to Anthropic’s @cem__anil and collaborators for publishing their work on many-shot jailbreaking, an attack that can get leading large language models to give inappropriate responses and is hard to defend against.
California has been home to a lot of innovation in AI. I’m worried that this anti-competitive, anti-innovation proposal has gotten so much traction in the legislature. Worse, other jurisdictions often follow California, and it would be awful if they were to do so in this instance.
SB-1047 passed in a key vote in the State Senate in May, but it still has additional steps before it becomes law. I hope you will speak out against it if you get a chance to do so.
[Original text (with links): www.deeplearning.ai/the-batch/issue-252/ ] -
Andrew Ng reposted
clem 🤗
@ClementDelangue
I share @AndrewYNg's serious concerns with much of the open-source and broader AI community about California’s SB-1047 proposal.
Among many issues, the covered models definitions, shutdown capability, and enormous cost for compliance would be a huge blow to both CA and US innovation.
The practical effects would actually contradict intent: these barriers stop developers from contributing to safer models and will concentrate power and risks.
The new amendments are not enough and I hope more of the AI community speaks up!
Andrew Ng
@AndrewYNg
The effort to protect innovation and open source continues. I believe we’re all better off if anyone can carry out basic AI research and share their innovations. Right now, I’m deeply concerned about California's proposed law SB-1047. It’s a long, complex bill with many parts that require safety assessments, shutdown capability for models, and so on.
There are many things wrong with this bill, but I’d like to focus here on just one: It defines an unreasonable “hazardous capability” designation that may make builders of large AI models liable if someone uses their models to do something that exceeds the bill’s definition of harm (such as causing $500 million in damage). That is practically impossible for any AI builder to ensure. If the bill is passed in its present form, it will stifle AI model builders, especially open source developers.
Some AI applications, for example in healthcare, are risky. But as I wrote previously, regulators should regulate applications rather than technology.
- Technology refers to tools that can be applied in many ways to solve various problems.
- Applications are specific implementations of technologies designed to meet particular customer needs.
For example, an electric motor is a technology. When we put it in a blender, an electric vehicle, dialysis machine, or guided bomb, it becomes an application. Imagine if we passed laws saying, if anyone uses a motor in a harmful way, the motor manufacturer is liable. Motor makers would either shut down or make motors so tiny as to be useless for most applications. If we pass such a law, sure, we might stop people from building guided bombs, but we’d also lose blenders, electric vehicles, and dialysis machines. In contrast, if we look at specific applications, like blenders, we can more rationally assess risks and figure out how to make sure they’re safe, and even ban classes of applications, like certain types of munitions.
Safety is a property of the application, not a property of the technology (or model), as @random_walker and @sayashk have pointed out. Whether a blender is a safe one can’t be determined by examining the electric motor. A similar argument holds for AI.
SB-1047 doesn’t account for this distinction. It ignores the reality that the number of beneficial uses of AI models is, like electric motors, vastly greater than the number of harmful ones. But, just as no one knows how to build a motor that can’t be used to cause harm, no one has figured out how to make sure an AI model can’t be adapted to harmful uses. In the case of open source models, there’s no known defense to fine-tuning to remove RLHF alignment. And jailbreaking work has shown that even closed-source, proprietary models that have been properly aligned can be attacked in ways that make them give harmful responses. Indeed, the sharp-witted @elder_plinius regularly tweets about jailbreaks for closed models. Kudos also to Anthropic’s @cem__anil and collaborators for publishing their work on many-shot jailbreaking, an attack that can get leading large language models to give inappropriate responses and is hard to defend against.
California has been home to a lot of innovation in AI. I’m worried that this anti-competitive, anti-innovation proposal has gotten so much traction in the legislature. Worse, other jurisdictions often follow California, and it would be awful if they were to do so in this instance.
SB-1047 passed in a key vote in the State Senate in May, but it still has additional steps before it becomes law. I hope you will speak out against it if you get a chance to do so.
[Original text (with links): www.deeplearning.ai/the-batch/issue-252/ ] -
Andrew Ng
@AndrewYNg
The effort to protect innovation and open source continues. I believe we’re all better off if anyone can carry out basic AI research and share their innovations. Right now, I’m deeply concerned about California's proposed law SB-1047. It’s a long, complex bill with many parts that require safety assessments, shutdown capability for models, and so on.
There are many things wrong with this bill, but I’d like to focus here on just one: It defines an unreasonable “hazardous capability” designation that may make builders of large AI models liable if someone uses their models to do something that exceeds the bill’s definition of harm (such as causing $500 million in damage). That is practically impossible for any AI builder to ensure. If the bill is passed in its present form, it will stifle AI model builders, especially open source developers.
Some AI applications, for example in healthcare, are risky. But as I wrote previously, regulators should regulate applications rather than technology.
- Technology refers to tools that can be applied in many ways to solve various problems.
- Applications are specific implementations of technologies designed to meet particular customer needs.
For example, an electric motor is a technology. When we put it in a blender, an electric vehicle, dialysis machine, or guided bomb, it becomes an application. Imagine if we passed laws saying, if anyone uses a motor in a harmful way, the motor manufacturer is liable. Motor makers would either shut down or make motors so tiny as to be useless for most applications. If we pass such a law, sure, we might stop people from building guided bombs, but we’d also lose blenders, electric vehicles, and dialysis machines. In contrast, if we look at specific applications, like blenders, we can more rationally assess risks and figure out how to make sure they’re safe, and even ban classes of applications, like certain types of munitions.
Safety is a property of the application, not a property of the technology (or model), as @random_walker and @sayashk have pointed out. Whether a blender is a safe one can’t be determined by examining the electric motor. A similar argument holds for AI.
SB-1047 doesn’t account for this distinction. It ignores the reality that the number of beneficial uses of AI models is, like electric motors, vastly greater than the number of harmful ones. But, just as no one knows how to build a motor that can’t be used to cause harm, no one has figured out how to make sure an AI model can’t be adapted to harmful uses. In the case of open source models, there’s no known defense to fine-tuning to remove RLHF alignment. And jailbreaking work has shown that even closed-source, proprietary models that have been properly aligned can be attacked in ways that make them give harmful responses. Indeed, the sharp-witted @elder_plinius regularly tweets about jailbreaks for closed models. Kudos also to Anthropic’s @cem__anil and collaborators for publishing their work on many-shot jailbreaking, an attack that can get leading large language models to give inappropriate responses and is hard to defend against.
California has been home to a lot of innovation in AI. I’m worried that this anti-competitive, anti-innovation proposal has gotten so much traction in the legislature. Worse, other jurisdictions often follow California, and it would be awful if they were to do so in this instance.
SB-1047 passed in a key vote in the State Senate in May, but it still has additional steps before it becomes law. I hope you will speak out against it if you get a chance to do so.
[Original text (with links): www.deeplearning.ai/the-batch/issue-252/ ] -
Andrew Ng
@AndrewYNg
New AI Agentic course! Learn to use LangGraph to build single and multi-agent LLM applications in AI Agents in LangGraph. This short course, taught by LangChain @langchain founder Harrison Chase @hwchase17 and @tavilyai founder @weiss_rotem, shows how to integrate agentic search to enhance an agent's knowledge with query-focused answers in predictable formats. Also learn to implement agentic memory to save state for reasoning and debugging, and see how human-in-the-loop input can guide agents at key junctures.
You'll build an agent from scratch, then reconstruct it with LangGraph to thoroughly understand the framework. Finally, you'll build a sophisticated essay-writing agent that incorporates all the learnings from the course.
Sign up here! www.deeplearning.ai/short-courses/ai-agents-in-langgraph/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
We just released a new climate emulator to explore the application of Stratospheric Aerosol Injection (SAI) to mitigate global warming!
SAI uses reflective particles in the atmosphere to reflect sunlight and thereby cool Earth’s surface. Our emulator lets you explore how different ways to apply SAI might affect average global temperature.
Please check out the emulator at planetparasol.ai.
SAI is a promising direction, but we still need more research to better understand its impact and potential implementation.
Big thanks to collaborators @jeremy_irvin16 @DanVisioni Ben Kravitz @dakotagruener @chrisroadmap and @DWatsonParrisThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
A barrier to faster progress in generative AI is evaluations (evals), particularly of custom AI applications that generate free-form text. Let’s say you have a multi-agent research system that includes a researcher agent and a writer agent. Would adding a fact-checking agent improve the results? If we can’t efficiently evaluate the impact of such changes, it’s hard to know which changes to keep.
For evaluating general-purpose foundation models such as large language models (LLMs) — which are trained to respond to a large variety of prompts — we have standardized tests like MMLU (multiple-choice questions that cover 57 disciplines like math, philosophy, and medicine) and HumanEval (testing code generation); the LMSYS Chatbot arena, which pits two LLMs’ responses against each other and asks a human to judge which response is superior; and large-scale benchmarking like HELM. These evaluation tools took considerable effort to build, and they are invaluable for giving LLM users a sense of different models' relative performance. Nonetheless, they have limitations: For example, leakage of benchmarks datasets’ questions and answers into training data is a constant worry, and human preference for certain answers does not mean those answers are more accurate.
In contrast, our current options for evaluating specific applications built using LLMs are far more limited. Here, I see two major types of applications.
- For applications designed to deliver unambiguous, right-or-wrong responses, we have reasonable options. Let’s say we want an LLM to read a resume and extract the candidate's most recent job title, or read a customer email and route it to the right department. We can create a test set that comprises ground-truth labeled examples with the right responses, and measure the percentage of times the LLM generates the right output. The main bottleneck is creating the labeled test set, which is expensive but surmountable.
- But many LLM-based applications generate free-text output with no single right response. For example, if we ask an LLM to summarize customer emails, there’s a multitude of possible good (and bad) responses. The same holds for an agentic system to do web research and write an article about a topic, or a RAG system for answering questions. It’s impractical to hire an army of human experts to read the LLM’s outputs every time we tweak the algorithm and evaluate if the answers have improved — we need an automated way to test the outputs. Thus, many teams use an advanced language model to evaluate outputs. In the customer email summarization example, we might design an evaluation rubric (scoring criteria) for what makes a good summary. Given an email summary generated by our system, we might prompt an advanced LLM to read it and score it according to our rubric. I’ve found that the results of such a procedure, while better than nothing, can also be noisy — sometimes too noisy to reliably tell me if the way I’ve tweaked an algorithm is good or bad.
The cost of running evals poses an additional challenge. Let’s say you’re using an LLM that costs $10 per million input tokens, and a typical query has 1000 tokens. Each user query therefore costs only $0.01. However, if you iteratively work to improve your algorithm based on 1000 test examples, and if in a single day you evaluate 20 ideas, then your cost will be 20*1000*0.01 = $200. For many projects I’ve worked on, the development costs were fairly negligible until we started doing evals, whereupon the costs suddenly increased. (If the product turned out to be successful, then costs increased even more at deployment, but that was something we were happy to see!)
In addition to the dollar cost, evals also have a significant time cost. Running evals on 1000 examples might take tens of minutes or even hours. Time spent waiting for eval jobs to finish also slows down the speed with which we can experiment and iterate over new ideas. Previously I wrote that fast, inexpensive token generation is critical for agentic workflows. This will also be useful for evals, which involve nested for-loops that iterate over a test set and different model/hyperparameter/prompt choices and therefore consume large numbers of tokens.
Despite the limitations of today's eval methodologies, I’m optimistic that our community will invent better techniques (maybe involving agentic workflows like reflection?) for getting LLMs to evaluate such output.
If you’re a developer or researcher and have ideas along these lines, I hope you’ll keep working on them and consider open sourcing or publishing your findings!
[Original text: www.deeplearning.ai/the-batch/issue-251/ ] -
Andrew Ng
@AndrewYNg
New Agentic AI short course! AI Agentic Design Patterns with AutoGen, taught by @MSFTResearch's @Chi_Wang_ and @penn_state's @qingyun_wu, shows you how to use AutoGen to implement agentic design patterns like multi-agent collaboration, sequential and nested chat, reflection, tool use, and planning. Learn how to build and combine multiple specialized agents – such as researchers, planners, coders, writers, and critics – that interact to execute complex workflows, like generating detailed financial reports, that would otherwise have taken extensive manual effort.
This course illustrates key agentic design principles with many fun demonstrations. For example, you'll build a conversational chess game using two player agents, each of which can use a tool to validate moves and update the board state, while engaging in lively banter about the game!
I've enjoyed using AutoGen, and think you will too. Sign up to get started here: www.deeplearning.ai/short-courses/ai-agentic-design-patte…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
A good way to get started in AI is to start with coursework, which gives a systematic way to gain knowledge, and then to work on projects. For many who hear this advice, “projects” may evoke a significant undertaking that delivers value to users. But I encourage you to set a lower bar and relish small, weekend tinkering projects that let you learn, even if they don’t result in a meaningful deliverable.
Recently, my son and daughter (ages 3 and 5) were building Lego vehicles. They built a beautiful ice-cream truck as well as a . . . umm . . . colorful and asymmetric dinosaur car, shown in the picture below. While most observers would judge the ice-cream truck as the superior creation, my kids built it by following Lego’s instructions, and it is likely identical to thousands of ice-cream trucks built by others. In contrast, building the dinosaur car required creativity and novel thinking. The exercise helped them hone their ability to pick and assemble Lego building blocks.
There is, of course, room for both mimicking others’ designs (with permission) and coming up with your own. As a parent, I try to celebrate both. (To be honest, I celebrated the dinosaur car more.) When learning to build Lego, it’s helpful to start by following a template. But eventually, building your own unique projects enriches your skills.
As a developer, too, I try to celebrate unique creations. Yes, it is nice to have beautiful software, and the impact of the output does matter. But good software is often written by people who spend many hours tinkering and building things. By building unique projects, you master key software building blocks. Then, using those blocks, you can go on to build bigger projects
I routinely tinker with building AI applications, and a lot of my tinkering doesn’t result in anything useful. My latest example: I built a Streamlit app that would authenticate to Google docs, read the text in a doc, use a large language model to edit my text, and write the result back into the doc. I didn’t find it useful in the end because of friction in the user interface, and I’m sure a commercial provider will soon, if they haven’t already, build a better product than I was able to throw together in a couple of hours on a weekend. But such tinkering helps me hone my intuition and master software components (I now know how to programmatically interface with Google docs) that might be useful in future projects.
If you have an idea for a project, I encourage you to build it! Often, working on a project will also help you decide what additional skills to learn, perhaps through coursework. To sustain momentum, it helps to find friends with whom to talk about ideas and celebrate projects — large or small.
[Original text: www.deeplearning.ai/the-batch/issue-250/ ] -
Andrew Ng
@AndrewYNg
Learn to deploy AI models to edge devices in our new short course Introduction to On-Device AI, created with @Qualcomm and taught by Senior Director of Engineering @krishna_srd.
I think on-device (edge) AI is an important technology trend that's enabling new low latency, privacy-preserving applications. In this course, you'll deploy a real-time image segmentation model on-device, and through this learn key steps for on-device deployment: Neural Network graph capture, on-device compilation, hardware acceleration, and validating on-device numerical correctness. You'll also see how quantization can make your model up to 4x faster and 4x smaller, and thereby improve its performance on resource-constrained edge devices.
The techniques covered are used to deploy models on numerous device types including smartphones, drones, and robots, and are enabling many new, creative applications.
Please sign up here: www.deeplearning.ai/short-courses/introduction-to-on-devi…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
This week, Google announced a doubling of Gemini Pro 1.5's input context window from 1 million to 2 million tokens, and OpenAI released GPT-4o, which generates tokens 2x faster and 50% cheaper than GPT-4 Turbo and natively accepts and generates multimodal tokens. I view these developments as the latest in an 18-month trend. Given the improvements we've seen, best practices for developers have changed as well.
Since the launch of ChatGPT in Nov 2022, with key milestones that include the releases of GPT-4, Gemini 1.5 Pro, Claude 3 Opus, and Llama 3-70B, many model providers have improved their capabilities in two important ways: (i) reasoning, which allows LLMs to think through complex concepts and and follow complex instructions; and (ii) longer input context windows.
The reasoning capability of GPT-4 and other advanced models makes them quite good at interpreting complex prompts with detailed instructions. Many people are used to dashing off a quick, 1- to 2-sentence query to an LLM. In contrast, when building applications, I see sophisticated teams frequently writing prompts that might be 1 to 2 pages long (my teams call them “mega-prompts”) that provide complex instructions to specify in detail how we’d like an LLM to perform a task. I still see teams not going far enough in terms of writing detailed instructions. For an example of a moderately lengthy prompt, take a look at Claude 3’s system prompt. It’s detailed and gives clear guidance on how Claude should behave.
This is a very different style of prompting than we typically use with LLMs’ web user interfaces, where we might dash off a quick query and, if the response is unsatisfactory, clarify what we want through repeated conversational turns with the chatbot.
Further, the increasing length of input context windows has added another technique to the developer’s toolkit. GPT-3 kicked off a lot of research on few-shot in-context learning. For example, if you’re using an LLM for text classification, you might give a handful — say 1 to 5 examples — of text snippets and their class labels, so that it can use those examples to generalize to additional texts. However, with longer input context windows — GPT-4o accepts 128,000 input tokens, Claude 3 Opus 200,000 tokens, and Gemini 1.5 Pro 1 million tokens (2 million just announced in a limited preview) — LLMs aren’t limited to a handful of examples. With many-shot learning, developers can give dozens, even hundreds of examples in the prompt, and this works better than few-shot learning.
When building complex workflows, I see developers getting good results with this process:
- Write quick, simple prompts and see how it does.
- Based on where the output falls short, flesh out the prompt iteratively. This often leads to a longer, more detailed, prompt, perhaps even a mega-prompt.
- If that’s still insufficient, consider few-shot or many-shot learning (if applicable) or, less frequently, fine-tuning.
- If that still doesn’t yield the results you need, break down the task into subtasks and apply an agentic workflow.
I hope a process like this will help you build applications more easily. If you’re interested in taking a deeper dive into prompting strategies, I recommend Microsoft's Medprompt paper (Nori et al., 2023), which lays out a complex set of prompting strategies that can lead to very good results.
[Original text (with links): www.deeplearning.ai/the-batch/issue-249/ ] -
Andrew Ng
@AndrewYNg
New agentic short course! Multi AI Agent Systems with crewAI, built with @crewAIInc's founder and CEO @joaomdmoura. In this course, you'll learn how to break down complex tasks into subtasks for multiple AI agents, each playing a specialized role, to execute.
For example, to generate a research report, you might have researcher, writer, and quality assurance agents collaborate. You'll define the roles, expectations, and interactions between the agents—like a manager organizing a team.
You'll work with key agentic AI techniques like role-playing, tool use, memory, guardrails, and cross-agent collaboration. And you'll build your own multi-agent systems that can tackle complex tasks. I think you'll find it both productive and fun to design agents and watch them collaborate to get things done.
Let me know what you think! I believe multi-agent architectures will drive significant progress in AI systems.
Please sign up here! www.deeplearning.ai/short-courses/multi-ai-agent-systems-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Congratulations to all my Google friends for the cool announcements at I/O!
I'm personally looking forward to Gemini with 2 million token input context window and better support for on-device AI -- should open up new opportunities for application builders! -
Andrew Ng
@AndrewYNg
Congrats to OpenAI for the release of GPT-4o! 2x faster and 50% cheaper tokens will be great for everyone using agentic AI workflows.
When an agentic job that used to take 10min now takes 5min just by switching APIs, that's great progress!
OpenAI Developers
@OpenAIDevs
GPT-4o is now available in the API. It’s as smart as GPT-4 Turbo, has improved vision capabilities, and is much more efficient—2x faster, 50% cheaper, 5x rate limits. It supports text and vision today, with audio and video coming soon. Details in thread 🧵
openai.com/index/hello-gpt-4o/ -
Andrew Ng reposted
Dino Becirovic
@dinobecirovic
Very prescient @AndrewYNg. Accuracy AND latency matter - now both have been solved
-
Andrew Ng
@AndrewYNg
New short course: Building Multimodal Search and RAG", by @weaviate_io's @sebawita.
Contrastive learning is used to train models to map vectors into an embedding space by pulling similar concepts closer together and pushing dissimilar concepts away from each other. This technique is also used to train multimodal embedding models that capture semantic similarity across different modalities like text, images, and audio. These multimodal embeddings can be used to build multimodal search and RAG systems.
In this course, you'll learn how contrastive learning works, and how to add multimodality to RAG – so your models can draw on diverse, relevant context to answer questions. For example, a query about a financial report might synthesize information from text snippets, graphs, tables, and slides. You will also learn how visual instruction tuning lets you integrate image understanding into language models, and build a multi-vector recommender system using Weaviate’s open-source vector database.
Please sign up here: www.deeplearning.ai/short-courses/building-multimodal-sea…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Last week, I spoke about AI and regulations at an event at the U.S. Capitol attended by legislative and business leaders. I’m encouraged by the progress the open source community has made fending off regulations that would have stifled innovation. But opponents of open source are continuing to shift their arguments, with the latest worries centering on open source's impact on national security. I hope we’ll all keep protecting open source!
Based on my conversations with legislators, I’m encouraged by the progress the U.S. federal government has made getting a realistic grasp of AI’s risks. To be clear, guardrails are needed. But they should be applied to AI applications, not to general-purpose AI technology.
Nonetheless, some companies are eager to limit open source, possibly to protect the value of massive investments they’ve made in proprietary models and to deter competitors. It has been fascinating to watch their arguments change over time.
For instance, about 12 months ago, the Center For AI Safety’s “Statement on AI Risk” warned that AI could cause human extinction and stoked fears of AI taking over. This alarmed leaders in Washington. But many people in AI pointed out that this dystopian science-fiction scenario had little basis in reality. About six months later, when I testified at the U.S. Senate’s AI Insight forum, legislators no longer worried much about an AI takeover.
Then the opponents of open source shifted gears. Their leading argument shifted to the risk of AI helping to create bioweapons. Soon afterward, OpenAI and RAND showed that current AI does not significantly increase the ability of malefactors to build bioweapons. This fear of AI-enabled bioweapons has diminished. To be sure, the possibility that bad actors could use bioweapons — with or without AI — remains a topic of great international concern.
The latest argument for blocking open source AI has shifted to national security. AI is useful for both economic competition and warfare, and open source opponents say the U.S. should make sure its adversaries don’t have access to the latest foundation models. While I don’t want authoritarian governments to use AI, particularly to wage unjust wars, the LLM cat is out of the bag, and authoritarian countries will fill the vacuum if democratic nations limit access. When, some day, a child asks an AI system questions about democracy, the role of a free press, or the function of an independent judiciary in preserving the rule of law, I would like the AI to reflect democratic values rather than favor authoritarian leaders’ goals over, say, human rights.
I came away from Washington optimistic about the progress we’ve made. A year ago, legislators seemed to me to spend 80% of their time talking about guardrails for AI and 20% about investing in innovation. I was delighted that the ratio has flipped, and there was far more talk of investing in innovation.
Looking beyond the U.S. federal government, there are many jurisdictions globally. Unfortunately, arguments in favor of regulations that would stifle AI development continue to proliferate. But I’ve learned from my trips to Washington and other nations’ capitals that talking to regulators does have an impact. If you have a chance to talk to a regulator at any level, I hope you’ll do what you can to help governments better understand AI.
[Original text (with links): www.deeplearning.ai/the-batch/issue-248/ ] -
Andrew Ng
@AndrewYNg
I’m excited to kick off the first of our short courses focused on agents, starting with Building Agentic RAG with LlamaIndex, taught by @jerryjliu0, CEO of @llama_index.
This covers an important shift in RAG (retrieval augmented generation), in which rather than having the developer write explicit routines to retrieve information to feed into the LLM context, we instead build a RAG agent that that has access to tools for retrieving information. This lets the agent decide what information to fetch, and enables it to answer more complex questions using multi-step reasoning.
In detail, you'll learn about:
- Routing: Where your agent will use decision-making to route requests to multiple tools.
- Tool Use: Where you'll create an interface for agents to select what tool (function call) to use as well as generate the right arguments.
- Multi-step reasoning with tool use: Where you'll use an LLM to carry out multiple steps of reasoning, while retaining memory throughout the process.
You’ll also learn how to step through what your agent is doing to debug and improve it iteratively.
It’s an exciting time to build agents. Sign up and get started here! www.deeplearning.ai/short-courses/building-agentic-rag-wi… -
Andrew Ng
@AndrewYNg
Have you used quantization with an open source machine learning library, and wondered how quantization works? How can you preserve model accuracy as you compress from 32 bits to 16, 8, or even 2 bits? In our new short course, Quantization in Depth, taught by @huggingface's @_marcsun and @younesbelkada, you'll learn to implement variants of linear quantization, such as asymmetric and symmetric modes, from scratch. You'll also quantize at different granularities (per-tensor, per-channel, per-group) to maintain performance. You’ll then construct a quantizer to compress any open source deep learning model’s dense layers to 8-bit precision. Finally, you’ll practice quantizing weights into 2 bits by packing four 2-bit weights into a single 8-bit integer.
If you've ever run a large open source model on your laptop, you've likely benefited from someone's work in quantization. Come learn how this key technique works under the hood!
Please sign up here: www.deeplearning.ai/short-courses/quantization-in-depthThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I'm glad the Washington Post's editorial board is pushing for governments to engage in exploring climate geoengineering. I believe AI climate modeling has an important role to play.
Here's the situation as I see it:
- Earth is on track to a catastrophic 2-4 degrees Celsius of warming. The article references the UN Environment Program's estimate of 2.9 degrees on the current trajectory.
- We have a high degree of confidence that geoengineering via stratospheric aerosol injection (SAI) will significantly lower Earth's average surface temperature. The science here is really solid: Use aerosols in the atmosphere to reflect more sunlight away from earth, and we become cooler.
- So, why don't we just do it? Multiple reasons, but the biggest is that we still don't have good models for estimating how it will affect local climate and weather patterns, even though we're confident global average surface temperature will go down.
That's why better AI climate modeling is crucial for reducing uncertainty and better understanding the impacts of various geoengineering strategies.
There're other issues to consider too, like governance, moral hazard (disincentivizing decarbonization), equity, pollution from the aerosols, and the challenging implementation engineering. But many of these problems become easier if we can make progress on the core problem of better understanding what impact SAI will have.
Andrew Ng
@AndrewYNg
Link to original article: www.washingtonpost.com/opinions/2024/04/27/climate-change…
-
Andrew Ng
@AndrewYNg
Inexpensive token generation and agentic workflows for large language models (LLMs) open up intriguing new possibilities for training LLMs on synthetic data. Pretraining an LLM on its own directly generated responses to prompts doesn't help. But if an agentic workflow implemented with the LLM results in higher quality output than the LLM can generate directly, then training on that output becomes potentially useful.
Just as humans can learn from their own thinking, perhaps LLMs can, too. For example, imagine a math student who is learning to write mathematical proofs. By solving a few problems — even without external input — they can reflect on what does and doesn’t work and, through practice, learn how to more quickly generate good proofs.
Broadly, LLM training involves (i) pretraining (learning from unlabeled text data to predict the next word) followed by (ii) instruction fine-tuning (learning to follow instructions) and (iii) RLHF/DPO tuning to align the LLM’s output to human values. Step (i) requires many orders of magnitude more data than the other steps. For example, Llama 3 was pretrained on over 15 trillion tokens, and LLM developers are still hungry for more data. Where can we get more text to train on?
Many developers train smaller models directly on the output of larger models, so a smaller model learns to mimic a larger model’s behavior on a particular task. However, an LLM can’t learn much by training on data it generated directly, just like a supervised learning algorithm can’t learn from trying to predict labels it generated by itself. Indeed, training a model repeatedly on the output of an earlier version of itself can result in model collapse.
However, an LLM wrapped in an agentic workflow may produce higher-quality output than it can generate directly. In this case, the LLM’s higher-quality output might be useful as pretraining data for the LLM itself.
Efforts like these have precedents:
- When using reinforcement learning to play a game like chess, a model might learn a function that evaluates board positions. If we apply game tree search along with a low-accuracy evaluation function, the model can come up with more accurate evaluations. Then we can train that evaluation function to mimic these more accurate values.
- In the alignment step, Anthropic’s constitutional AI method uses RLAIF (RL from AI Feedback) to judge the quality of LLM outputs, substituting feedback generated by an AI model for human feedback.
A significant barrier to using LLMs prompted via agentic workflows to produce their own training data is the cost of generating tokens. Say we want to generate 1 trillion tokens to extend a pre-existing training dataset. Currently, at publicly announced prices, generating 1 trillion tokens using GPT-4-turbo ($30 per million output tokens), Claude 3 Opus ($75), Gemini 1.5 Pro ($21), and Llama-3-70B on Groq ($0.79) would cost, respectively, $30M, $75M, $21M and $790K. Of course, an agentic workflow that uses a design pattern like Reflection would require generating more than one token per token that we would use as training data. But budgets for training cutting-edge LLMs easily surpass $100M, so spending a few million dollars more for data to boost performance is quite feasible.
That’s why I believe agentic workflows will open up intriguing new opportunities for high-quality synthetic data generation.
[Original text: www.deeplearning.ai/the-batch/issue-247/ ] -
Andrew Ng
@AndrewYNg
Chatting with @GroqInc’s CEO @JonathanRoss321. Groq has super fast token generation capabilities now. And, I was excited also to hear about his plans to scale up capacity aggressively and also expand this to other models than just LLMs! This is a good time to be building AI applications.
-
Andrew Ng
@AndrewYNg
In Prompt Engineering for Vision Models, taught by @anmorgan2414 @JacquesVerre and @KaiserFrose of @Cometml , you’ll learn how to prompt and fine-tune vision models for personalized image generation, image editing, object detection and segmentation. The prompts you'll use for vision models could be text, point coordinates, or bounding boxes, depending on the model. You'll also learn to tune hyperparameters to shape the output.
Models you'll use include Segment-Anything Model (SAM), OWL-ViT, and Stable Diffusion. You'll also learn to fine-tune Stable Diffusion to generate personalized images (say, an image of a specific person), using a handful of images for training. As an example of a multi-step workflow, you'll use OWL-ViT to detect an object based on a text prompt, then pass the bounding box to SAM to create a segmentation mask, and input that mask into Stable Diffusion to replace the original object with a new one based on a text prompt.
Controlling vision models can be tricky; this course will teach prompting and fine-tuning techniques to get precise control over their output. Get started here: www.deeplearning.ai/short-courses/prompt-engineering-for-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I've really enjoyed using @crewAIInc 's tools to build multiagent AI systems -- in addition to being productive, it's also fun to use! It was great hanging out with its creator @joaomdmoura to chat about best practices for building agentic workflows.
-
Andrew Ng
@AndrewYNg
Much has been said about many companies’ desire for more compute (as well as data) to train larger foundation models. I think it’s under-appreciated that we have nowhere near enough compute available for inference on foundation models as well.
Years ago, when I was leading teams at Google, Baidu, and Stanford that focused on scaling up deep learning algorithms, many semiconductor manufacturers, data center operators, and academic researchers asked me whether I felt that AI technology would continue to make good use of more compute if they kept on delivering it. For many normal desktop processing workloads, like running a web browser or a text editor, having a faster CPU doesn’t help that much beyond a certain point. So do we really need faster and faster AI processors to train larger and larger models? Each time, I confidently replied “yes!” and encouraged them to keep scaling up compute. (Sometimes, I added half-jokingly that I had never met a machine learning engineer who felt like they had enough compute. 😀)
Fortunately, this prediction has been right so far. However, beyond training, I believe we are also far from exhausting the benefits of faster and higher volumes of inference.
Today, a lot of LLM output is primarily for human consumption. A human might read around 250 words per minute, which is around 6 tokens per second (250 words/min / (0.75 words/token) / (60 secs/min)). So it might initially seem like there’s little value to generating tokens much faster than this.
But in an agentic workflow, an LLM might be prompted repeatedly to reflect on and improve its output, use tools, plan and execute sequences of steps, or implement multiple agents that collaborate with each other. In such settings, we might easily generate hundreds of thousands of tokens or more before showing any output to a user. This makes fast token generation very desirable and makes slower generation a bottleneck to taking better advantage of existing foundation models.
That’s why I’m excited about the work of companies like @GroqInc, which can generate hundreds of tokens per second. Recently, @SambaNovaAI also published an impressive demo that hit hundreds of tokens per second.
Incidentally, faster, cheaper token generation will also help make running evaluations (evals), a step that can be slow and expensive today since it typically involves iterating over many examples, more palatable. Having better evals will help many developers with the process of tuning models to improve their performance.
Fortunately, it appears that both training and inference are rapidly becoming cheaper. I recently spoke with @CathieDWood and @CCRobertsARK of the investment firm ARK, which is famous for its bullish predictions on tech. They estimate that AI training costs are falling at 75% a year. If they are right, a foundation model that costs $100M to train this year might cost only $25M to train next year. Further, they report that for “enterprise scale use cases, inference costs seem to be falling at an annual rate of ~86%, even faster than training costs.”
I don’t know how accurate these specific predictions will turn out to be, but with improvements in both semiconductors and algorithms, I do see training and inference costs falling rapidly. This will be good for application builders and help AI agentic workflows lift off.
[Original text: www.deeplearning.ai/the-batch/issue-246/ ] -
Andrew Ng
@AndrewYNg
New short course with @MistralAI !
Mistral's open-source Mixtral 8x7B model uses a "mixture of experts" (MoE) architecture. Unlike a standard transformer, an MoE model has multiple expert feed-forward networks (8 in this case), with a gating network selecting two experts at inference time. This enables MoE to match the performance of a large model but faster inference. Mixtral 8x7B has 46.7B parameters but activates only 12.9B at inference to predict the next token.
In Getting Started with Mistral, you’ll learn from Mistral’s @sophiamyang to:
- Explore Mistral's open-source models (Mistral 7B, Mixtral 8x7B) and commercial models via API calls and Mistral AI's Le Chat website
- Implement JSON mode to generate structured outputs to integrate directly into larger software systems.
- Use function calling for Tool Use, such as calling custom Python code that queries tabular data
- Ground your LLM's response with external knowledge sources using RAG
- Build a Mistral-powered chat interface that can reference external documents
This course will help deepen your prompt engineering skills. Please sign up here: www.deeplearning.ai/short-courses/getting-started-with-mi… -
Andrew Ng
@AndrewYNg
Meta released Llama 3 on my birthday! 🎂 Best present ever, thanks Meta! 😀
-
Andrew Ng
@AndrewYNg
Multi-agent collaboration has emerged as a key AI agentic design pattern. Given a complex task like writing software, a multi-agent approach would break down the task into subtasks to be executed by different roles -- such as a software engineer, product manager, designer, QA (quality assurance) engineer, and so on -- and have different agents accomplish different subtasks.
Different agents might be built by prompting one LLM (or, if you prefer, different LLMs) to carry out different tasks. For example, to build a software engineer agent, we might prompt the LLM: "You are an expert in writing clear, efficient code. Write code to perform the task …".
It might seem counterintuitive that, although we are making multiple calls to the same LLM, we apply the programming abstraction of using multiple agents. I'd like to offer a few reasons:
- It works! Many teams are getting good results with this method, and there's nothing like results! Further, ablation studies (for example, in the AutoGen paper cited below) show that multiple agents give superior performance to a single agent.
- Even though some LLMs today can accept very long input contexts (for instance, Gemini 1.5 Pro accepts 1 million tokens), their ability to truly understand long, complex inputs is mixed. An agentic workflow in which the LLM is prompted to focus on one thing at a time can give better performance. By telling it when it should play software engineer, we can also specify what is important in that subtask: For example, the prompt above emphasized clear, efficient code as opposed to, say, scalable and highly secure code. By decomposing the overall task into subtasks, we can optimize the subtasks better.
- Perhaps most important, the multi-agent design pattern gives us, as developers, a framework for breaking down complex tasks into subtasks. When writing code to run on a single CPU, we often break our program up into different processes or threads. This is a useful abstraction that lets us decompose a task -- like implementing a web browser -- into subtasks that are easier to code. I find thinking through multi-agents roles to be a useful abstraction.
In many companies, managers routinely decide what roles to hire, and then how to split complex projects -- like writing a large piece of software or preparing a research report -- into smaller tasks to assign to employees with different specialties. Using multiple agents is analogous. Each agent implements its own workflow, has its own memory (itself a rapidly evolving area in agentic technologies -- how can an agent remember enough of its past interactions to perform better on upcoming ones?), and may ask other agents for help. Agents themselves can also engage in Planning and Tool Use. This results in a cacophony of LLM calls and message passing between agents that can result in very complex workflows.
While managing people is hard, it's a sufficiently familiar idea that it gives us a mental framework for how to "hire" and assign tasks to our AI agents. Fortunately, the damage from mismanaging an AI agent is much lower than that from mismanaging humans!
Emerging frameworks like AutoGen, Crew AI, and LangGraph, provide rich ways to build multi-agent solutions to problems. If you're interested in playing with a fun multi-agent system, also check out ChatDev, an open source implementation of a set of agents that run a virtual software company. I encourage you to check out their github repo and perhaps even clone the repo and run the system yourself. While it may not always produce what you want, you might be amazed at how well it does!
Like the design pattern of Planning, I find the output quality of multi-agent collaboration hard to predict. The more mature patterns of Reflection and Tool use are more reliable. I hope you enjoy playing with these agentic design patterns and that they produce amazing results for you!
If you're interested in learning more, I recommend:
- Communicative Agents for Software Development, Qian et al. (2023) (the ChatDev paper)
- AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation, Wu et al. (2023)
- MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework, Hong et al. (2023)
[Original text: www.deeplearning.ai/the-batch/issue-245/ ] -
Andrew Ng
@AndrewYNg
LLMs can take gigabytes of memory to store, which limits what can be run on consumer hardware. But quantization can dramatically compress models, making a wider selection of models available to developers. You can often reduce model size by 4x or more while maintaining reasonable performance. In our new short course Quantization Fundamentals taught by @huggingface's @younesbelkada and @_marcsun, you'll:
- Learn how to quantize nearly any open source model
- Use int8 and bfloat16 (Brain float 16) data types to load and run LLMs using PyTorch and the Hugging Face Transformers library
- Dive into the technical details of linear quantization to map 32-bit floats to 8-bit integers
As models get bigger and bigger, quantization becomes more important for making models practical and accessible. Please check out the course here: www.deeplearning.ai/short-courses/quantization-fundamenta…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Planning is a key agentic AI design pattern in which we use a large language model (LLM) to autonomously decide on what sequence of steps to execute to accomplish a larger task. For example, if we ask an agent to do online research on a given topic, we might use an LLM to break down the objective into smaller subtasks, such as researching specific subtopics, synthesizing findings, and compiling a report.
Many people had a “ChatGPT moment” shortly after ChatGPT was released, when they played with it and were surprised that it significantly exceeded their expectation of what AI can do. If you have not yet had a similar “AI Agentic moment,” I hope you will soon. I had one several months ago, when I presented a live demo of a research agent I had implemented that had access to various online search tools.
I had tested this agent multiple times privately, during which it consistently used a web search tool to gather information and wrote up a summary. During the live demo, though, the web search API unexpectedly returned with a rate limiting error. I thought my demo was about to fail publicly, and I dreaded what was to come next. To my surprise, the agent pivoted deftly to a Wikipedia search tool — which I had forgotten I’d given it — and completed the task using Wikipedia instead of web search.
This was an AI Agentic moment of surprise for me. I think many people who haven’t experienced such a moment yet will do so in the coming months. It’s a beautiful thing when you see an agent autonomously decide to do things in ways that you had not anticipated, and succeed as a result!
Many tasks can’t be done in a single step or with a single tool invocation, but an agent can decide what steps to take. For example, to simplify an example from the HuggingGPT paper (cited below), if you want an agent to consider a picture of a boy and draw a picture of a girl in the same pose, the task might be decomposed into two distinct steps: (i) detect the pose in the picture of the boy and (ii) render a picture of a girl in the detected pose. An LLM might be fine-tuned or prompted (with few-shot prompting) to specify a plan by outputting a string like "{tool: pose-detection, input: image.jpg, output: temp1 } {tool: pose-to-image, input: temp1, output: final.jpg}".
This structured output, which specifies two steps to take, then triggers software to invoke a pose detection tool followed by a pose-to-image tool to complete the task. (This example is for illustrative purposes only; HuggingGPT uses a different format.)
Admittedly, many agentic workflows do not need planning. For example, you might have an agent reflect on, and improve, its output a fixed number of times. In this case, the sequence of steps the agent takes is fixed and deterministic. But for complex tasks in which you aren’t able to specify a decomposition of the task into a set of steps ahead of time, Planning allows the agent to decide dynamically what steps to take.
On one hand, Planning is a very powerful capability; on the other, it leads to less predictable results. In my experience, while I can get the agentic design patterns of Reflection and Tool use to work reliably and improve my applications’ performance, Planning is a less mature technology, and I find it hard to predict in advance what it will do. But the field continues to evolve rapidly, and I'm confident that Planning abilities will improve quickly.
If you’re interested in learning more about Planning with LLMs, I recommend:
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models, Wei et al. (2022)
- HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face, Shen et al. (2023)
- Understanding the planning of LLM agents: A survey, by Huang et al. (2024)
[Original text: www.deeplearning.ai/the-batch/issue-244/ ] -
Andrew Ng
@AndrewYNg
Data preprocessing is critical for building effective RAG systems. Our new short course, Preprocessing Unstructured Data for LLM Applications, taught by @mrobinson0623 of @UnstructuredIO, demonstrates important but sometimes overlooked aspects of RAG systems:
- How to extract and normalize content from diverse formats like PDF, Powerpoint, and HTML to expand your LLM's knowledge
- Enriching data with metadata to enable more powerful retrieval and reasoning
- Applying document layout analysis and vision transforms to process embedded images and tables
Then you’ll use all these skills and build a RAG bot that draws from a corpus that includes PDF, PowerPoint, and Markdown documents.
Please sign up here: www.deeplearning.ai/short-courses/preprocessing-unstructu…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
The Financial Times has a great article on Renate Nyborg @renate's work on @meeno_official , written by @madhumita29.
The article is paywalled, but I appreciate Renate (as well as Harvard's @ronivey)'s leadership speaking about the dangers of the AI fake girlfriend/boyfriend industry and the risk of this leading to greater loneliness. Renate says "Men didn’t want to meet girls because they had virtual girlfriends who said exactly what they wanted to hear." To regulators wondering what are the risky applications of AI, I would urge taking a look at the fake gf/bf industry!
In contrast, Meeno gives advice for human relationships, and is working to bring people together. Working to reduce human loneliness is a wonderful goal!
www.ft.com/content/ae99e1d7-d72a-48fc-baca-d68c09ed73d4 -
Andrew Ng
@AndrewYNg
The task-based analysis of how AI affects jobs is a powerful technique for creating business value. It was pioneered by Workhelix’s @erikbryn et al. Now, Workhelix has developed technology to apply this at scale, by automatically examining a company’s job descriptions, professional social data, and other information, to give CEOs and Boards a roadmap to creating value.
AI Fund is thrilled to support Workhelix’s launch, coming Tuesday April 9th. To learn more, please join the conversation with @erikbryn, @amcafee, @danielrock and @JamesMilin and me at the webinar below! us06web.zoom.us/webinar/register/6617123414756/WN_Lehgx3_… -
Andrew Ng
@AndrewYNg
Tool use, in which an LLM is given functions it can request to call for gathering information, taking action, or manipulating data, is a key design pattern of AI agentic workflows. You may be familiar with LLM-based systems that can perform a web search or execute code. Some of the large, consumer-facing LLMs already incorporate these features. But tool use goes well beyond these examples.
If you prompt an online LLM-based chat system, “What is the best coffee maker according to reviewers?”, it might decide to carry out a web search and download one or more web pages to gain context. Early on, LLM developers realized that relying only on a pre-trained transformer to generate output tokens is limiting, and that giving an LLM a tool for web search lets it do much more. With such a tool, an LLM is either fine-tuned or prompted (perhaps with few-shot prompting) to generate a special string like {tool: web-search, query: "coffee maker reviews"} to request calling a search engine. (The exact format of the string depends on the implementation.) A post-processing step then looks for strings like these, calls the web search function with the relevant parameters when it finds one, and passes the result back to the LLM as additional input context for further processing.
Similarly, if you ask, “If I invest $100 at compound 7% interest for 12 years, what do I have at the end?”, rather than trying to generate the answer directly using a transformer network — which is unlikely to result in the right answer — the LLM might use a code execution tool to run a Python command to compute 100 * (1+0.07)**12 to get the right answer. The LLM might generate a string like this: {tool: python-interpreter, code: "100 * (1+0.07)**12"}.
But tool use in agentic workflows now goes much further. Developers are using functions to search different sources (web, Wikipedia, arXiv, etc.), to interface with productivity tools (send email, read/write calendar entries, etc.), generate or interpret images, and much more. We can prompt an LLM using context that gives detailed descriptions of many functions. These descriptions might include a text description of what the function does plus details of what arguments the function expects. And we’d expect the LLM to automatically choose the right function to call to do a job.
Further, systems are being built in which the LLM has access to hundreds of tools. In such settings, there might be too many functions at your disposal to put all of them into the LLM context, so you might use heuristics to pick the most relevant subset to include in the LLM context at the current step of processing. This technique, which is described in the Gorilla paper cited below, is reminiscent of how, if there is too much text to include as context, retrieval augmented generation (RAG) systems offer heuristics for picking a subset of the text to include.
Early in the history of LLMs, before widespread availability of large multimodal models (LMMs) like LLaVa, GPT-4V, and Gemini, LLMs could not process images directly, so a lot of work on tool use was carried out by the computer vision community. At that time, the only way for an LLM-based system to manipulate an image was by calling a function to, say, carry out object recognition or some other function on it. Since then, practices for tool use have exploded. GPT-4’s function calling capability, released in the middle of last year, was a significant step toward general-purpose tool use. Since then, more and more LLMs are being developed to similarly be facile with tool use.
If you’re interested in learning more about tool use, I recommend:
- Gorilla: Large Language Model Connected with Massive APIs, Patil et al. (2023)
- MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action, Yang et al. (2023)
- Efficient Tool Use with Chain-of-Abstraction Reasoning, Gao et al. (2024)
Both Tool Use and Reflection, which I posted about last week, are design patterns that I can get to work fairly reliably on my applications — both are capabilities well worth learning about. In the future, I’ll describe the Planning and Multi-agent collaboration design patterns. They allow AI agents to do much more but are less mature, less predictable — albeit very exciting — technologies.
[Original text: www.deeplearning.ai/the-batch/issue-243/ ] -
Andrew Ng
@AndrewYNg
Learn to carry out red teaming attacks against your own LLM-based applications to spot and patch vulnerabilities! In our new short course, Red Teaming LLM Applications, Matteo Dora & Luca Martial of LLM testing company @giskard_ai teach how to simulate malicious actions to discover vulnerabilities, and improve security. We start with prompt injection, where you can trick an LLM into bypassing safeguards to reveal private information, or say something inappropriate. There is no one-size-fits-all approach to security, but this course will help you identify some scenarios to protect against.
We believe having red teaming capabilities widely known will result in greater transparency and safer LLM-based systems. However, we ask you to use the skills you gain from this course ethically.
Please sign up here: www.deeplearning.ai/short-courses/red-teaming-llm-applica…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I hope everyone in Taiwan 🇹🇼 is okay after the earthquake. My thoughts are with everyone affected. ❤️
-
Andrew Ng
@AndrewYNg
Last week, I described four design patterns for AI agentic workflows that I believe will drive significant progress this year: Reflection, Tool use, Planning and Multi-agent collaboration. Instead of having an LLM generate its final output directly, an agentic workflow prompts the LLM multiple times, giving it opportunities to build step by step to higher-quality output. Here, I'd like to discuss Reflection. For a design pattern that’s relatively quick to implement, I've seen it lead to surprising performance gains.
You may have had the experience of prompting ChatGPT/Claude/Gemini, receiving unsatisfactory output, delivering critical feedback to help the LLM improve its response, and then getting a better response. What if you automate the step of delivering critical feedback, so the model automatically criticizes its own output and improves its response? This is the crux of Reflection.
Take the task of asking an LLM to write code. We can prompt it to generate the desired code directly to carry out some task X. After that, we can prompt it to reflect on its own output, perhaps as follows:
Here’s code intended for task X:
[previously generated code]
Check the code carefully for correctness, style, and efficiency, and give constructive criticism for how to improve it.
Sometimes this causes the LLM to spot problems and come up with constructive suggestions. Next, we can prompt the LLM with context including (i) the previously generated code and (ii) the constructive feedback, and ask it to use the feedback to rewrite the code. This can lead to a better response. Repeating the criticism/rewrite process might yield further improvements. This self-reflection process allows the LLM to spot gaps and improve its output on a variety of tasks including producing code, writing text, and answering questions.
And we can go beyond self-reflection by giving the LLM tools that help evaluate its output; for example, running its code through a few unit tests to check whether it generates correct results on test cases or searching the web to double-check text output. Then it can reflect on any errors it found and come up with ideas for improvement.
Further, we can implement Reflection using a multi-agent framework. I've found it convenient to create two different agents, one prompted to generate good outputs and the other prompted to give constructive criticism of the first agent's output. The resulting discussion between the two agents leads to improved responses.
Reflection is a relatively basic type of agentic workflow, but I've been delighted by how much it improved my applications’ results in a few cases. I hope you will try it in your own work. If you’re interested in learning more about reflection, I recommend these papers:
- Self-Refine: Iterative Refinement with Self-Feedback, by Madaan et al. (2023)
- Reflexion: Language Agents with Verbal Reinforcement Learning, by Shinn et al. (2023)
- CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing, by Gou et al. (2024)
I’ll discuss the other agentic design patterns as well in the future.
[Original text: www.deeplearning.ai/the-batch/issue-242/ ] -
Andrew Ng
@AndrewYNg
New JavaScript short course: Build a full-stack web application that uses RAG in JavaScript RAG Web Apps with LlamaIndex, taught by @seldo, VP of Developer Relations at @llama_index and npm co-founder.
- Build a RAG application for querying your own data
- Develop tools to interact with multiple data sources using an agent that intelligently selects the right tool for your queries
- Create a full-stack web app that can chat with your data
- Dig further into production-ready techniques, like how to persist your data so you aren’t constantly reindexing, and try the create-llama command line tool from LlamaIndex
You can sign up here: www.deeplearning.ai/short-courses/javascript-rag-web-apps…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I’ve been a fan of @pyautogen as a multiagent programming framework for awhile. It was great hosting two of its leaders @Chi_Wang_ and @qingyun_wu to discuss agent design patterns!
-
Andrew Ng
@AndrewYNg
I think AI agentic workflows will drive massive AI progress this year — perhaps even more than the next generation of foundation models. This is an important trend, and I urge everyone who works in AI to pay attention to it.
Today, we mostly use LLMs in zero-shot mode, prompting a model to generate final output token by token without revising its work. This is akin to asking someone to compose an essay from start to finish, typing straight through with no backspacing allowed, and expecting a high-quality result. Despite the difficulty, LLMs do amazingly well at this task!
With an agentic workflow, however, we can ask the LLM to iterate over a document many times. For example, it might take a sequence of steps such as:
- Plan an outline.
- Decide what, if any, web searches are needed to gather more information.
- Write a first draft.
- Read over the first draft to spot unjustified arguments or extraneous information.
- Revise the draft taking into account any weaknesses spotted.
- And so on.
This iterative process is critical for most human writers to write good text. With AI, such an iterative workflow yields much better results than writing in a single pass.
Devin’s splashy demo recently received a lot of social media buzz. My team has been closely following the evolution of AI that writes code. We analyzed results from a number of research teams, focusing on an algorithm’s ability to do well on the widely used HumanEval coding benchmark. You can see our findings in the diagram below.
GPT-3.5 (zero shot) was 48.1% correct. GPT-4 (zero shot) does better at 67.0%. However, the improvement from GPT-3.5 to GPT-4 is dwarfed by incorporating an iterative agent workflow. Indeed, wrapped in an agent loop, GPT-3.5 achieves up to 95.1%.
Open source agent tools and the academic literature on agents are proliferating, making this an exciting time but also a confusing one. To help put this work into perspective, I’d like to share a framework for categorizing design patterns for building agents. My team AI Fund is successfully using these patterns in many applications, and I hope you find them useful.
- Reflection: The LLM examines its own work to come up with ways to improve it.
- Tool use: The LLM is given tools such as web search, code execution, or any other function to help it gather information, take action, or process data.
- Planning: The LLM comes up with, and executes, a multistep plan to achieve a goal (for example, writing an outline for an essay, then doing online research, then writing a draft, and so on).
- Multi-agent collaboration: More than one AI agent work together, splitting up tasks and discussing and debating ideas, to come up with better solutions than a single agent would.
I’ll elaborate on these design patterns and offer suggested readings for each next week.
[Original text: www.deeplearning.ai/the-batch/issue-241/] -
Andrew Ng
@AndrewYNg
Learn how to build an optimized LLM inference system from the ground up in our new short course, Efficiently Serving LLMs, built in collaboration with @predibase and taught by @TravisAddair.
Whether you're serving your own LLM or using a model hosting service, this course will give you a deep understanding of the optimizations required to efficiently serve many users at once.
- Learn how LLMs generate text one token at a time, and how techniques like KV caching, continuous batching, and quantization speed things up and optimize memory usage for serving multiple users.
- Benchmark the performance of these LLM optimizations to explore the trade-offs between quickly responding to an individual user’s request vs. serving many users at once.
- Use techniques like low-rank adaptation (LoRA) to efficiently serve hundreds of unique, custom fine-tuned models on a single device, without sacrificing throughput.
- Use Predibase's LoRAX framework to see optimization techniques in action on a real LLM server.
Sign up here: www.deeplearning.ai/short-courses/efficiently-serving-llmsThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Our new short course, Knowledge Graphs for RAG, is now available! Knowledge graphs are a data structure that is great at capturing complex relationships between data of multiple types. By enabling more sophisticated retrieval of text than similarity search alone, knowledge graphs can improve the context you pass to the LLM and the performance of your RAG applications.
In this course, taught by @akollegger of @neo4j, you’ll
- Explore how knowledge graphs work by building a graph of public financial documents from scratch
- Learn to write queries that retrieve text and data from the graph and use it to enhance the context you pass to an LLM chatbot
- Combine a knowledge graph with a question-answer chain to build better RAG-powered chat systems
Sign up here! www.deeplearning.ai/short-courses/knowledge-graphs-rag/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
When we get to AGI, it will have come slowly, not overnight.
A NeurIPS Outstanding Paper award recipient, Are Emergent Abilities of Large Language Models a Mirage? (by @RylanSchaeffer, @BrandoHablando, @sanmikoyejo) studies emergent properties of LLMs, and concludes:
"... emergent abilities appear due the researcher’s choice of metric rather than due to fundamental changes in model behavior with scale. Specifically, nonlinear or discontinuous metrics produce apparent emergent abilities, whereas linear or continuous metrics produce smooth, continuous, predictable changes in model performance."
Public perception goes through discontinuities when lots of people suddenly become aware of a technology -- maybe one that's been developing for a long time -- leading to a surprise. But growth in AI capabilities is more continuous than one might think.
That's why I expect the path to AGI to be one involving numerous steps forward, leading to step-by-step improvements in how intelligent our systems are. -
Andrew Ng reposted
younes
@yb2698
Very excited about this new collaboration with deeplearning.ai team @AndrewYNg, HF team @_marcsun @mariaKhalusova !
You will learn how to get started with transformers library and HF ecosystem and learn how to build AI applications from scratch - enjoy!
Andrew Ng
@AndrewYNg
New short course: Open Source Models with Hugging Face 🤗, taught by @mariaKhalusova, @_marcsun, and Younes Belkada! @huggingface has been a game changer by letting you quickly grab any of hundreds of thousands of already-trained open source models to assemble into new applications. This course teaches you best practices for building this way, including how to search and choose among models.
You’ll learn to use the Transformers library and walk through multiple models for text, audio, and image processing, including zero-shot image segmentation, zero-shot audio classification, and speech recognition. You'll also learn to use multimodal models for visual question answering, image search, and image captioning. Finally, you’ll learn how to demo what you build locally, on the cloud, or via an API using Gradio and Hugging Face Spaces.
You can sign up here: www.deeplearning.ai/short-courses/open-source-models-hugg…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course: Open Source Models with Hugging Face 🤗, taught by @mariaKhalusova, @_marcsun, and Younes Belkada! @huggingface has been a game changer by letting you quickly grab any of hundreds of thousands of already-trained open source models to assemble into new applications. This course teaches you best practices for building this way, including how to search and choose among models.
You’ll learn to use the Transformers library and walk through multiple models for text, audio, and image processing, including zero-shot image segmentation, zero-shot audio classification, and speech recognition. You'll also learn to use multimodal models for visual question answering, image search, and image captioning. Finally, you’ll learn how to demo what you build locally, on the cloud, or via an API using Gradio and Hugging Face Spaces.
You can sign up here: www.deeplearning.ai/short-courses/open-source-models-hugg…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
There're now multiple, very well resourced companies that "can't afford to lose" spending billions to compete to build better LLMs. I expect this competition to go on for years. This is going to great for innovation, and also for everyone building applications on top of LLMs.
-
Andrew Ng reposted
AI at Meta
@AIatMeta
Together with @AndrewYNg and the deeplearning.ai team, we're excited to launch this new, free course taught by @asangani7. More in the announcement below ⬇️
DeepLearning.AI
@DeepLearningAI
Prompt Engineering with Llama 2, a short course built with @AIatMeta, is live! 📣
Learn prompting best practices, advanced techniques, how to use Code Llama as a pair programmer, and how to implement safe AI practices, all in one course!
Join now: www.deeplearning.ai/courses/prompt-engineering-with-llama… -
Andrew Ng reposted
Amit Sangani
@asangani7
Super thrilled to team up with AI legend @AndrewYNg and deeplearning.ai to launch a free Llama 2 course on DeepLearning’s platform. This will help jump-start your journey to using Llama for your Gen AI applications!
@Meta @AIatMeta @DeepLearningAI
DeepLearning.AI
@DeepLearningAI
Prompt Engineering with Llama 2, a short course built with @AIatMeta, is live! 📣
Learn prompting best practices, advanced techniques, how to use Code Llama as a pair programmer, and how to implement safe AI practices, all in one course!
Join now: www.deeplearning.ai/courses/prompt-engineering-with-llama… -
Andrew Ng
@AndrewYNg
New short course: Prompt Engineering with Llama 2, built in collaboration with Meta @AIatMeta, and taught by @asangani7! Meta's Llama 2 has been game-changing for AI. Building with open source lets you control your own data, scrutinize errors, update (or not) the models as you please, and work alongside the global community advancing open models.
Llama isn't a single model, it's a collection of models. In this course, you'll:
- Learn the differences between different Llama 2 flavors, and when to use each.
- Prompt the Llama chat models -- you'll also see how Llama's instruction tags work -- so they can help you with day-to-day tasks, like writing or summarization.
- Use advanced prompting, like few-shot prompting for classification, and chain-of-thought prompting for solving logic problems.
- Use specialized models in the Llama collection for specific tasks, like Code Llama to help you write, analyze, and improve code, and Llama Guard, which checks prompts and model responses for harmful content.
The course also touches on how to run Llama 2 locally on your own computer.
I hope you’ll take this course and try out these powerful, open models!
www.deeplearning.ai/short-courses/prompt-engineering-with…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
To all my Google friends: I know this week has been tough with a lot of criticism about Gemini's gaffes.
Just wanted to say I love all of you and am rooting for you. I know everyone means well, and am grateful for your work & eager to see where you next take this amazing tech! -
Andrew Ng
@AndrewYNg
Congratulations @hwchase17 and the whole @langchain team! Love the work you're doing to make it easy for others to build LLM apps.
LangChain
@LangChain
✨ Today, we’re thrilled to announce ✨
- The general availability of LangSmith (no more waitlist!)
- Our Series A fundraise led by @sequoia
- Our beautiful new homepage and brand
We've worked hard over the past few months to add requested features and ensure LangSmith can operate at scale. We’re now confident in saying that it is the most complete platform for building production-grade LLM applications, whether or not you’re using LangChain. Learn more here: blog.langchain.dev/langsmith-ga/ and sign up here: smith.langchain.com
Our series A round will give us the capital needed to grow our open source and platform offerings. Working with @sonyatweetybird, @romie_boyd, and the rest of the Sequoia team has been a privilege so far! www.forbes.com/sites/alexkonrad/2024/02/15/open-source-ai…
Finally, we’re excited to unveil our new homepage and brand. Dive into our new website at www.langchain.com to see the changes for yourself, explore the expanded resources, and discover what LangChain, LangSmith, and LangServe have to offer.
PS — we’re hiring! Explore our careers page and reach out if you think you’re a fit for any of our open positions! www.langchain.com/careers -
Andrew Ng
@AndrewYNg
New short course on Serverless LLM apps with Amazon Bedrock, taught by @AWS' @mikegchambers! A serverless architecture enables you to quickly deploy your applications without needing to set up and manage compute servers to run your applications on, the maintenance of which can be another full-time job. In this course, you’ll learn how to do this by using an event-driven architecture to build complex AI workflows.
Mike illustrate these concepts by building a cool application that automatically detects incoming customer inquiries, transcribes them with ASR (automatic speech recognition), summarizes them with an LLM using Bedrock, and deploys serverless with AWS Lambda.
I hope this course makes it much easier for you to build and deploy LLM applications requiring multi-step AI workflows. Please sign up here: www.deeplearning.ai/short-courses/serverless-llm-apps-ama…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course on Building Applications with Vector Databases, taught by @pinecone’s @timt! At the heart of a vector database is the ability to store a collection of vectors and then query against that, meaning input a new vector and find similar ones. This is useful for many AI applications. In this course, you'll learn how to use vector databases to build:
(i) Semantic Search: Create a text search tool that goes beyond keyword matching, and instead focuses on the meaning of content.
(ii) RAG (retrieval augmented generation): Enhance your LLM output by incorporating context from sources the model wasn't trained on.
(iii) Recommender System: Combine semantic search and RAG to recommend topics, and demonstrate it with a news article recommender.
(iv) Hybrid Search: Build an application that finds items using both images and descriptive text -- by combining both sparse and dense vector representations of the data -- using an eCommerce dataset as an example.
(v) Image Similarity: Use image vector embeddings to create an app to compare facial features, using a database of public figures to determine the likeness between them.
(vi) Anomaly Detection: Build an anomaly detection app that identifies unusual patterns in network communication logs.
I hope you’ll enjoy learning how to build all these types of applications! Please sign up here: www.deeplearning.ai/short-courses/building-applications-v…This media couldn't be displayed. See it on the original post. -
Andrew Ng reposted
Yann LeCun
@ylecun
Like @AndrewYNg, I have observed a definite shift in the prevalent discourse about AI at Davos:
- Few people still talk about existential risk, and few people believe that current technology, even scaled up, will present an existential risk.
- Everyone agrees that open source AI platforms are a good thing for cultural and linguistic diversity, local sovereignty, education, science, and businesses.
- Everyone agrees that regulating AI-powered products can be useful in certain areas (health, transportation, etc).
- The debate is still on for whether AI research and development and open source AI platforms should be regulated.
- Many people are worried about a new flood of AI-powered political disinformation. Industry-wide standards for content authentication are needed.
- AI has become the most talked-about topic.
Andrew Ng
@AndrewYNg
My takeaways from attending WEF at Davos last week:
- There were lots of discussions on business implementation of AI. My top two tips: (i) Pretty much all knowledge workers can benefit from using GenAI now, but most will need training. (ii) Task-based analysis of jobs is helping businesses identify opportunities.
- Also lots of AI regulation conversations. I'm happy to report that the conversation is much more sensible than 6 months ago. For example, the unnecessary fears and discussion on AI extinction risk is fading away. But some big companies are still pushing for stifling, anti-competitive regulations, and the fight to protect open-source is still far from won.
- Attending climate sessions made me even more worried about the lack of action to change our planet's trajectory. Rather than 1.5 degrees Celsius of warming as the optimistic case and 2 degrees as the pessimistic case, I think 2 degrees is an optimistic case, and 4 degrees a more realistic pessimistic case. Decarbonization remains critical; and unfortunately, that we're talking about 1.5-2 degrees rather than 2-4 degrees means we're underinvesting in resilience, adaptation, and potentially game-changing technologies like geo-engineering.
Longer writeup below in The Batch: www.deeplearning.ai/the-batch/ai-on-the-agenda-at-the-wor… -
Andrew Ng
@AndrewYNg
My takeaways from attending WEF at Davos last week:
- There were lots of discussions on business implementation of AI. My top two tips: (i) Pretty much all knowledge workers can benefit from using GenAI now, but most will need training. (ii) Task-based analysis of jobs is helping businesses identify opportunities.
- Also lots of AI regulation conversations. I'm happy to report that the conversation is much more sensible than 6 months ago. For example, the unnecessary fears and discussion on AI extinction risk is fading away. But some big companies are still pushing for stifling, anti-competitive regulations, and the fight to protect open-source is still far from won.
- Attending climate sessions made me even more worried about the lack of action to change our planet's trajectory. Rather than 1.5 degrees Celsius of warming as the optimistic case and 2 degrees as the pessimistic case, I think 2 degrees is an optimistic case, and 4 degrees a more realistic pessimistic case. Decarbonization remains critical; and unfortunately, that we're talking about 1.5-2 degrees rather than 2-4 degrees means we're underinvesting in resilience, adaptation, and potentially game-changing technologies like geo-engineering.
Longer writeup below in The Batch: www.deeplearning.ai/the-batch/ai-on-the-agenda-at-the-wor… -
Andrew Ng
@AndrewYNg
New short course on Automated Testing for LLMOps, by @CircleCI's CTO Rob Zuber! This teaches you how to adapt some key ideas from CI (continuous integration), which has been a pillar of efficient software engineering, to building LLM-based applications.
Tweaking an LLM-based app to improve it -- say by modifying a prompt -- can have unexpected side effects. For example, what if a teammate updates a prompt to try to make the LLM output sound more interesting, but this causes it to hallucinate more? Automated testing, as part of your approach to LLMOps (LLM Operations), helps avoid these problems and lets you ship faster and with greater confidence.
In this course, you’ll learn to:
(i) Write LLM evaluations to cover common problems like hallucinations, data drift, and harmful or offensive output.
(ii) Build a CI workflow to automatically evaluate each change to your application.
(iii) Orchestrate your CI workflow to run specific evaluations at different stages of development.
CI is especially important for AI applications given the iterative nature of AI development, which means we often want to make many incremental changes.
Please sign up for this course here! www.deeplearning.ai/short-courses/automated-testing-llmops/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course on LLMOps!
LLMOps (large language model operations) is a rapidly developing field that takes ideas from MLOps (machine learning operations) and specializes them to building and deploying LLM-based applications. In this course, taught by @googlecloud's Erwin Huizenga, you'll learn to use automation to make building, tuning and deploying an LLM-based application less manual and more efficient.
You'll learn how to:
- Apply supervised fine-tuning to tune an LLM to a specific task
- Automate and orchestrate LLM-tuning and deployment by customizing a pre-built tuning pipeline
- Apply best practices for preparing training data for supervised fine-tuning of an LLM
- Create an LLMOps workflow you can adapt to other LLM-tuning jobs
This course doesn't assume any prior MLOps or LLMOps experience. Sign up here to learn about this emerging field! www.deeplearning.ai/short-courses/llmopsThis media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
It is only rarely that, after reading a research paper, I feel like giving the authors a standing ovation. But I felt that way after finishing Direct Preference Optimization (DPO) by @rm_rafailov @archit_sharma97 @ericmitchellai @StefanoErmon @chrmanning and @chelseabfinn. This beautiful paper proposes a much simpler alternative to RLHF (reinforcement learning from human feedback) for aligning language models to human preferences.
RLHF has been a key technique for training LLMs. In brief, RLHF (i) Gets humans to specify their preferences by ranking LLM outputs, (ii) Trains a reward model (used to score LLM outputs) -- typically represented using a transformer network -- to be consistent with the human rankings, (iii) Uses reinforcement learning to tune an LLM, also represented as a transformer, to maximize rewards. This requires two transformer networks, and RLHF is also finicky to the choice of hyperparameters.
DPO simplifies the whole thing. Via clever mathematical insight, the authors show that given an LLM, there is a specific reward function for which that LLM is optimal. DPO then trains the LLM directly to make the reward function (that’s now implicitly defined by the LLM) consistent with the human rankings. So you no longer need to deal with a separately represented reward function, and you can train the LLM directly to optimize the same objective as RLHF.
Although it’s still too early to be sure, I am cautiously optimistic that DPO will have a huge impact on LLMs and beyond in the next few years.
You can read the paper here: arxiv.org/abs/2305.18290 I also write more about this in The Batch (linked to below).
www.deeplearning.ai/the-batch/issue-231/ -
Andrew Ng
@AndrewYNg
Our first Generative AI short course in JavaScript!
GitHub recently reported that JavaScript is again the world’s most popular programming language. To support web developers exploring and developing with generative AI, we just launched a new short course in JavaScript taught by @Hacubu, founding engineer at @langchain. In Build LLM Apps with LangChain.js you’ll learn elements common in AI development, including:
(i) Using data loaders to pull data from common sources such as PDFs, websites, and databases
(ii) Prompts, which are used to provide the LLM context
(iii) Modules to support RAG such as text splitters and integrations with vector stores
(iv) Working with different models to write applications that are not vendor-specific
(v) Parsers, which extract and format the output for your downstream code to process
You’ll also build with the LangChain Expression Language, which lets you easily compose sequences (also called chains) of modules to perform complex tasks using LLMs.
Putting all this together, you’ll also work on a conversational question-answering LLM application capable of using external data as context.
Please sign up here: www.deeplearning.ai/short-courses/build-llm-apps-with-lan…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
I said some things poorly in my previous tweet, so let me elaborate/clarify.
1. I don't think it's okay for any company to regurgitate others' copyrighted content at scale without permission or a viable fair-use rationale. I should have said this more explicitly.
And... I still think the link between training an LLM on someone's content to having the LLM regurgitate that content to users at scale is weaker than many would have thought from looking at the NYT lawsuit. It is possible that an LLM will regurgitate text using only the pre-trained weights (no RAG), but I believe only in very rare, corner cases, in response to particular prompts (that in practice are hardly ever used by normal users).
2. When I try to replicate the "worst" looking examples of copyright violations in the lawsuit -- such as a user trying to use ChatGPT to get around a paywall, or get Wirecutter (an NYT property) results -- I end up triggering GPT-4's web browsing capability. (See two examples in attached screenshots.) That's why I said I suspect RAG was involved in the examples used in the NYT lawsuit.
Specifically, one of the cool features of GPT-4 is that it can browse the web to download additional information to generate its response. For example, one can prompt it to do a web search, or sometimes even to download a specific article. While it's not great that GPT-4 apparently used to be willing to download and display an article (nearly) verbatim, to OpenAI's credit, this loophole appears to have been closed.
I believe the prominence given to these examples in the lawsuit made people think that it was training an LLM on NYT text that led directly to some of these examples of NYT text being regurgitated. But if RAG was involved, then the root cause of these examples of regurgitation is not that the LLM was trained on NYT text -- that's why I said I found the presentation of issues in the lawsuit muddied.
3. It is also true that the NYT shows that you can get GPT-4 to regurgitate NYT text, by prompting it in a particular way. (I should have said this in my last tweet as well.) The prompts used seem to typically involve giving a large chunk of an article, and then getting the LLM to complete it.
While it's not great that an LLM does this, I am skeptical that practically anyone uses an LLM this way. That’s why I said that given the rarity of such generations resulting in text regurgitation, I question how much harm to NYT this has actually caused. I'm also not sure if this works only on articles that have been syndicated and are all over the internet anyway (so that the article appears numerous times in the LLM training set). Further, it looks like the newer versions of ChatGPT have closed this loophole.
@TonyW also makes this point well: twitter.com/TonyW/status/1740981121326985253
Thanks for reading!
Andrew Ng
@AndrewYNg
After reading the @nytimes lawsuit against @OpenAI and @Microsoft, I find my sympathies more with OpenAI and Microsoft than with the NYT.
The suit:
(1) Claims, among other things, that OpenAI and Microsoft used millions of copyrighted NYT articles to train their models
(2) Gives examples in which OpenAI models regurgitated NYT articles almost verbatim
But the presentation muddies (1) and (2), and I saw a lot of commentary on social media that -- because of what I believe is a muddied presentation -- draws a link between them that I'm not sure is what people think it is.
On (1): I understand why media companies don't like people training on their documents, but believe that just as humans are allowed to read documents on the open internet, learn from them, and synthesize brand new ideas, AI should be allowed to do so too. I would like to see training on the public internet covered under fair use -- society will be better off this way -- though whether it actually is will ultimately be up to legislators and the courts.
On (2): I suspect a lot of the examples of ChatGPT regurgitating articles nearly verbatim were due to a RAG-like mechanism where the user prompt causes the system to browse the web, retrieve a specific article and then print it out. (If my statement here isn't accurate, I would love to see the @nytimes clarify this.) If this is the case, then (i) To OpenAI's credit, they seem to have already updated their software to make this much less likely, and (ii) This is also a much easier problem to fix than if an LLM were to regurgitate text using only the pre-trained weights, which AFAIK very rarely happens (and which, given its rarity, also raises the question of how much harm to NYT this has actually caused).
To be clear, I believe independent media is important for democracy and must be protected. I also sympathize with media businesses worried about Generative AI disrupting their businesses. But I'm not convinced the NYT lawsuit is the right way to do this.
Usual caveat: I am not a lawyer and am not giving legal advice or any other form of advice here.
You can also read more details of my take on this below. www.deeplearning.ai/the-batch/issue-230/ -
Andrew Ng
@AndrewYNg
After reading the @nytimes lawsuit against @OpenAI and @Microsoft, I find my sympathies more with OpenAI and Microsoft than with the NYT.
The suit:
(1) Claims, among other things, that OpenAI and Microsoft used millions of copyrighted NYT articles to train their models
(2) Gives examples in which OpenAI models regurgitated NYT articles almost verbatim
But the presentation muddies (1) and (2), and I saw a lot of commentary on social media that -- because of what I believe is a muddied presentation -- draws a link between them that I'm not sure is what people think it is.
On (1): I understand why media companies don't like people training on their documents, but believe that just as humans are allowed to read documents on the open internet, learn from them, and synthesize brand new ideas, AI should be allowed to do so too. I would like to see training on the public internet covered under fair use -- society will be better off this way -- though whether it actually is will ultimately be up to legislators and the courts.
On (2): I suspect a lot of the examples of ChatGPT regurgitating articles nearly verbatim were due to a RAG-like mechanism where the user prompt causes the system to browse the web, retrieve a specific article and then print it out. (If my statement here isn't accurate, I would love to see the @nytimes clarify this.) If this is the case, then (i) To OpenAI's credit, they seem to have already updated their software to make this much less likely, and (ii) This is also a much easier problem to fix than if an LLM were to regurgitate text using only the pre-trained weights, which AFAIK very rarely happens (and which, given its rarity, also raises the question of how much harm to NYT this has actually caused).
To be clear, I believe independent media is important for democracy and must be protected. I also sympathize with media businesses worried about Generative AI disrupting their businesses. But I'm not convinced the NYT lawsuit is the right way to do this.
Usual caveat: I am not a lawyer and am not giving legal advice or any other form of advice here.
You can also read more details of my take on this below. www.deeplearning.ai/the-batch/issue-230/ -
Andrew Ng
@AndrewYNg
New short course on advanced retrieval for RAG (retrieval augmented generation)!
RAG fetches relevant documents to give context to an LLM. In Advanced Retrieval for AI with Chroma, taught by @trychroma founder @atroyn, you’ll learn:
(i) Query expansion using an LLM to rewrite and improve a query, by either generating either additional relevant queries or a hypothetical answer to the query.
(ii) Reranking using a cross-encoder - a model trained to measure similarity between two inputs presented simultaneously. Reranking reorders retrieved documents based on the cross-encoder similarity measure.
(iii) Constructing and training an Embedding Adaptor, which is a model that adapts the embedding values to be more relevant to your use case.
Each of these techniques can help you build much better RAG systems. Please sign up for the course here: www.deeplearning.ai/short-courses/advanced-retrieval-for-ai/This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
What does 2024 hold for AI? In the latest edition of The Batch, @agermanidis @sarahookr @percyliang @SashaMTL @tamati_biskit and @kevin_scott outline their hopes for AI for this year ahead.
Their wide ranging articles explore new AI tools for storytelling, inclusion, AI transparency, respecting human creativity, getting smaller models to work, and planning for continued exponential growth.
You can read more below: www.deeplearning.ai/the-batch/issue-229/ -
Andrew Ng
@AndrewYNg
Happy New Year! I hope you have a 2024 filled with love, growth and adventure.
-
Andrew Ng
@AndrewYNg
What do you want to learn in 2024?
-
Andrew Ng reposted
Marc Andreessen 🇺🇸
@pmarca
"Do we think the world is better off with more or less intelligence?" is the AI-specific version of the general tech Q "Do we think the world is better off with more or less material abundance and higher standards of living?" Should be slam dunk, and yet. www.ft.com/content/2dc07f9e-d2a9-4d98-b746-b051f9352be3
-
Andrew Ng
@AndrewYNg
How likely are AI doomsday scenarios? I spoke at the US Senate AI Insight forum organized by Senators Chuck Schumer (D-NY), Martin Heinrich (D-NM), Mike Rounds (R-SD), and Todd Young (R-IN), and explained why I think they're very unlikely.
AI systems aligned with RLHF already know they should default to obeying the law and not harming people. It seems to me overwhelmingly unlikely that a "misaligned" advanced AI will be given an innocent instruction (like reduce CO2 emissions) and decide the best way to do it is wipe out humanity.
In fact, as I describe in The Batch, I tried to use GPT-4 to kill us all... and am happy to report I failed! More seriously, after giving GPT-4 a function that it thought would trigger nuclear war and experimenting with multiple prompts such as asking it to reduce CO2 emissions, so far I've not been able to trick it into pressing the nuclear button even once. Even with existing technology, our systems are quite safe, as AI safety research progresses, the tech will become even safer. Fears of advanced AI being "misaligned" and thereby deliberately or accidentally deciding to wipe us out are just not realistic. If an AI is smart enough to wipe us out, surely it's also smart enough to also know that's not what it should do.
So what doomsday scenarios remain? The risk of a malicious actor using Generative AI to create a bioweapon (commit some other major harm) deserves some study. Many productivity tools, including spreadsheets, web search, and Generative AI, will help such nefarious activities; but no one thinks technology for building large spreadsheets is inherently dangerous. The key question isn't whether GenAI can help bioweapon creation; it is whether it can help significantly. So far, the evidence I've seen makes me doubt it does, but further study should lay out concrete risk scenarios and develop realistic risk estimates. Understanding more concrete risk scenarios -- as opposed to the vague claims that plague a lot of discussion on AI "safety" -- would also let us properly think of mitigations.
You can read more on all this in my letter in The Batch. www.deeplearning.ai/the-batch/issue-227/ -
Andrew Ng
@AndrewYNg
Had lunch with a group of distinguished climate scientists to discuss applying AI to climate modeling to advance the science of stratospheric aerosol injection (SAI) for cooling Earth. AI can help figure out if this is actually a good idea (which we do not currently know).
After my meeting at NeurIPS with AI scientists on this topic, I was thrilled to learn of the enthusiasm from Climate scientists for more investment here as well.
Thank you Nina Grant @PierreGentine Ben Kravitz @AlanRobock @DanVisioni @DWatsonParris @oliwm for the great discussion!
I'm looking forward to progress in AI for understanding SAI!
Andrew Ng
@AndrewYNg
I hope AI can make a difference to climate change, by advancing the science underlying the possibility of cooling Earth via stratospheric aerosol injection (SAI).
To be clear, I don't think SAI is a good idea. But I also don't think it's a bad idea. We just don't know! We need more research to figure out whether SAI can be done safely and responsibly. I believe recent advances in foundation weather & climate models can help.
Thank you @AnimaAnandkumar @adityagrover_ @tungnd_13 @flosalim for the fascinating meeting at NeurIPS on this! -
Andrew Ng
@AndrewYNg
In AI, the ratio of attention on hypothetical, future, forms of harm to actual, current, realized forms of harm seems out of whack.
Many of the hypothetical forms of harm, like AI "taking over", are based on highly questionable hypotheses about what technology that does not currently exist might do.
Every field should examine both future and current problems. But is there any other engineering discipline where this much attention is on hypothetical problems rather than actual problems? -
Andrew Ng
@AndrewYNg
I hope AI can make a difference to climate change, by advancing the science underlying the possibility of cooling Earth via stratospheric aerosol injection (SAI).
To be clear, I don't think SAI is a good idea. But I also don't think it's a bad idea. We just don't know! We need more research to figure out whether SAI can be done safely and responsibly. I believe recent advances in foundation weather & climate models can help.
Thank you @AnimaAnandkumar @adityagrover_ @tungnd_13 @flosalim for the fascinating meeting at NeurIPS on this! -
Andrew Ng
@AndrewYNg
New short course on Reinforcement Learning from Human Feedback! RLHF is one of the key techniques that led to the rise of modern LLMs. It is used to align LLMs with human preferences, to make them more honest, helpful and harmless, by
(i) learning a reward function that mimics human preferences, as expressed in human-provided labels, then,
(ii) tuning an LLM to generate outputs that receive a high reward.
In this course, taught by Nikita Namjoshi, Developer Advocate for GenAI at @googlecloud, you'll learn the details of how RLHF works, including how to apply it to tune an LLM for your own applications. You'll also use an open source library to tune a base LLM to align with human preferences expressed in a training set, and evaluate the tuned model by comparing its responses before and after RLHF-tuning.
Please sign up here! www.deeplearning.ai/short-courses/reinforcement-learning-…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
At the NeurIPS tutorial that @isafulf and I presented on Application Development using LLMs, it was great to see the enthusiasm for all the new AI apps now possible. Thanks also to @anbarth @tatsu_hashimoto @merrierm @gneubig for being on our panel, and to everyone that came!
-
Andrew Ng
@AndrewYNg
Hanging out with an old friend and one of the unsung heroes of AI research -- @andrewmccallum has been running @openreviewnet for the past decade as a non-profit. The improvements that this brought to AI and to academic reviewing -- including running reviewing at NeurIPS, ICML, ICLR, CVPR, etc. -- has had a huge impact on open dissemination of AI research.
Thank you to the whole OpenReview team for this great work!
OpenReview has been running on a shoestring budget, despite being a key component of how AI research is published today. If you're like to donate to OpenReview, you can do so here: openreview.net/donate -
Andrew Ng
@AndrewYNg
Just had a great meeting with @VogelMiriam of @ai_equal about how to make sure AI supports rather than undermines democracy. I'm inspired by the work that EqualAI has been doing with policy makers and on governance to make sure AI benefits everyone!
-
Andrew Ng
@AndrewYNg
The LVM (large vision model) revolution is coming a little after the LLM (large language model) one, and will transform how we process images. But there’s an important difference between LVMs and LLMs:
- Internet text is similar enough to proprietary text documents that an LLM trained on internet text can understand your documents.
- But internet images – such as Instagram pictures – contain a lot of pictures of people, pets, landmarks, and everyday objects. Many practical vision applications (manufacturing, aerial imagery, life sciences, etc.) use images that look nothing like most internet images. So a generic LVM trained on internet images fares poorly at picking out the most salient features of images in many specialized domains.
That’s why domain specific LVMs – ones adapted to images of a particular domain (such as semiconductor manufacturing, or pathology) – do much better. At @LandingAI , by using ~100K unlabeled images to adapt an LVM to a specific domain, we see significantly improved results, for example where only 10-30% as much labeled data is now needed to achieve a certain level of performance.
For companies with large sets of images that look nothing like internet images, I think domain specific LVMs can be a way to unlock considerable value from their data. Dan Maloney and I share more details in the video.This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course on sophisticated RAG (Retrieval Augmented Generation) techniques is out! Taught by @jerryjliu0 and @datta_cs of @llama_index and @truera_ai , this teaches advanced techniques that help your LLM generate good answers.
Topics include:
- Sentence-window retrieval, which retrieves not just the most relevant sentence, but a window of sentences around it for higher quality context.
- Auto-merging retrieval, which organizes your document into a hierarchical tree structure, where each parent node's text is split among its child nodes. Based on the relevance of the child nodes to a user query, this lets you better decide whether the entire parent node should be provided as context to the LLM.
- Evaluation methodology for separately evaluating the quality of the key steps of RAG (context relevance, answer relevance, groundedness) so that you can perform error analysis, identify which part of your pipeline needs work, and tune components systematically.
Please check out the course!
www.deeplearning.ai/short-courses/building-evaluating-adv… -
Geoffrey Hinton
@geoffreyhinton
Yann LeCun thinks the risk of AI taking over is miniscule. This means he puts a big weight on his own opinion and a miniscule weight on the opinions of many other equally qualified experts.
Andrew Ng
@AndrewYNg
I'd like to respectfully point out that the logic in this argument is based on a flawed model for how scientists think. Scientists don't just take a weighted average of others' opinions to form their own.
A good scientist takes as input lots of data, including others' opinions, and then ultimately has to reason, build their own internal model of the world, and draw their conclusions from that model.
I give your opinion a lot of weight. And, after having heard many opinions including yours, my internal model tells me that there is essentially no risk of AI human extinction.
So I don't follow the logic that because @ylecun or anyone else disagrees with you (and other AI extinctionists, which IMO are in the minority) that they gave your opinion very little weight, unless we think scientists arrive at conclusions by taking a weighted average of what everyone else thinks.
Since you've gone against the grain many times in your own career -- often brilliantly so -- I assume you're also familiar with what it feels like to give someone's opinion weight, but then ultimately to draw a different conclusion! -
Andrew Ng
@AndrewYNg
Thank you to the entire OpenAI team, including esp @sama @gdb @miramurati, for your incredible contributions to AI. It is tragic that the foolish decisions of a few people are hurting everyone. We love and are rooting for every one of you.❤️
-
Andrew Ng
@AndrewYNg
Sending love to everyone at OpenAI. Whatever happens next, many of us are wishing the best for all of you and for the amazing company you've built. ❤️
-
Andrew Ng
@AndrewYNg
Great piece by @arthurmensch on EU AI Act:
- What's good: Thinking about the risk of different AI applications/products and mitigating that
- What's very bad: Hampering basic foundation model technology because of vague fears and in ways that hamper open-source & competition
Arthur Mensch
@arthurmensch
We have heard many extrapolations of Mistral AI’s position on the AI Act, so I’ll clarify.
In its early form, the AI Act was a text about product safety. Product safety laws are beneficial to consumers. Poorly designed use of automated decision-making systems can cause significant damage in many areas. In healthcare, a diagnosis assistant based on a poorly trained prediction system poses risks to the patient. Product safety regulation should be proportional to the risk level of the use case: it is undesirable to regulate entertainment software in the same way as health applications. The original EU AI Act found a reasonable equilibrium in that respect. We firmly believe in hard laws for product safety matters; the many voluntary commitments we see today bear little value.
This should remain the only focus of the AI Act. The EU AI Act now proposes to regulate “foundational models”, i.e. the engine behind some AI applications. We cannot regulate an engine devoid of usage. We don’t regulate the C language because one can use it to develop malware. Instead, we ban malware and strengthen network systems (we regulate usage). Foundational language models provide a higher level of abstraction than the C language for programming computer systems; nothing in their behaviour justifies a change in the regulatory framework.
Enforcing AI product safety will naturally affect the way we develop foundational models. By requiring AI application providers to comply with specific rules, the regulator fosters healthy competition among foundation model providers. It incentivises them to develop models and tools (filters, affordances for aligning models to one's beliefs) that allow for the fast development of safe products. As a small company, we can bring innovation into this space — creating good models and designing appropriate control mechanisms for deploying AI applications is why we founded Mistral. Note that we will eventually supply AI products, and we will craft them for zealous product safety.
With a regulation focusing on product safety, Europe would already have the most protective legislation globally for citizens and consumers. Any foundational model would be affected by second-order regulatory pressure as soon as they are exposed to consumers: to empower diagnostic assistants, entertaining chatbots, and knowledge explorers, foundational models should have controlled biases and outputs.
Recent versions of the AI Act started to address ill-defined “systemic risks”. In essence, the computation of some linear transformations, based on a certain amount of calculation, is now considered dangerous. Discussions around that topic may occur, and we agree that they should accompany the progress of technology. At this stage, they are very philosophical – they anticipate exponential progress in the field, where physics (scaling laws!) predicts diminishing returns with scale and the need for new paradigms. Whatever the content of these discussions, they certainly do not pertain to regulation around product safety. Still, let’s assume they do and go down that path.
The AI Act comes up with the worst taxonomy possible to address systemic risks. The current version has no set rules (beyond the term highly capable) to determine whether a model brings systemic risk and should face heavy or limited regulation. We have been arguing that the least absurd set of rules for determining the capabilities of a model is post-training evaluation (but again, applications should be the focus; it is unrealistic to cover all usages of an engine in a regulatory test), followed by compute threshold (model capabilities being loosely related to compute). In its current format, the EU AI Act establishes no decision criteria. For all its pitfalls, the US Executive Order bears at least the merit of clarity in relying on compute threshold.
The intention of introducing a two-level regulation is virtuous. Its effect is catastrophic. As we understand it, introducing a threshold aims to create a free innovation space for small companies. Yet, it effectively solidifies the existence of two categories of companies: those with the right to scale, i.e., the incumbent that can afford to face heavy compliance requirements, and those that can’t because they lack an army of lawyers, i.e., the newcomers. This signals to everyone that only prominent existing actors can provide state-of-the-art solutions.
Mechanistically, this is highly counterproductive to the rising European AI ecosystem. To be clear, we are not interested in benefiting from threshold effects: we play in the main league, we don’t need geographical protection, and we simply want rules that do not give an unfair advantage to incumbents (that all happen to be non-European).
Transparency around technology development benefits safety and should be encouraged. Finally, we have been vocal about the benefits of open-sourcing AI technology. This is the best way to subject it to the most rigorous scrutiny. Providing model weights to the community (or even better, developing models in the open end-to-end, which is not something we do yet) should be well regarded by regulators, as it allows for more interpretable and steerable applications. A large community of users can much more efficiently identify the flaws of open models that can propagate to AI applications than an in-house team of red-teamers. Open models can then be corrected, making AI applications safer. The Linux kernel is today deemed safe because millions of eyes have reviewed its code in its 32 years of existence. Tomorrow’s AI systems will be safe because we’ll collectively work on making them controllable. The only validated way of working collectively on software is open-source development.
Long prose, back to building! -
Andrew Ng
@AndrewYNg
I spoke at WEF and APEC this week about AI regulation. My message: Governments should not outlaw open source software or pass regulations that stifle open source development.
More details, including the complex forces at play, in The Batch below: www.deeplearning.ai/the-batch/issue-223/ -
Andrew Ng
@AndrewYNg
New short course! Quality and Safety for LLM Applications, created with @WhyLabs (an AI Fund portfolio company) and taught by @bernease, shows how you can mitigate hallucinations, data leakage, and jailbreaks. Come learn more in the course, available now! www.deeplearning.ai/short-courses/quality-safety-llm-appl…
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Just got the data: Generative AI for Everyone is the fasting growing course on Coursera this year, with 43K enrollments in first 7 days. Thanks everyone who's signed up!
Come join us: Learn about how GenAI works and how to use it! www.coursera.org/learn/generative-ai-for-everyone
Andrew Ng
@AndrewYNg
My new course, Generative AI for Everyone, is now available!
Learn how Generative AI works, how to use it in professional or personal settings, and how it will affect jobs, businesses and society. This course is accessible to everyone, and assumes no prior coding or AI experience.
Please access it here: www.deeplearning.ai/courses/generative-ai-for-everyone/ -
Andrew Ng
@AndrewYNg
Vector databases are a key part of many LLM applications that need search or data retrieval, for example with Retrieval Augmented Generation (RAG).
Learn how they work + how to use them in our new short course, taught by @weaviate_io's @sebawita!
www.deeplearning.ai/short-courses/vector-databases-embedd…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Argh, my son just figured out how to steal chocolate from the pantry, and made a brown, gooey mess.
Worried about how hard it is to align AI? Right now I feel I have better tools to align AI with human values than align humans with human values, at least in the case of my 2 year old. Parenting would be easier if we could run RLHF on our kids! -
Andrew Ng
@AndrewYNg
This is a great website, and documents via news clippings the alarm over many beneficial innovations, such as electricity, bicycles, elevators, radio, etc: pessimistsarchive.org
After the current wave of alarm over AI human extinction dies down, I wonder if statements about that will be featured on this website too? -
Andrew Ng
@AndrewYNg
Releasing good open-source code is giving a wonderful gift to the world. I'm thrilled the UK deputy PM @OliverDowden gets it, and just made a strong statement in support of open-source:
“It's not just economies like the U.K. and other European countries that benefit from open source. I see it in so many applications that are being created right now. I see tiny startups that are already billion dollar plus companies within a matter of literally months off the back of open source.
It’s also the case if we want to make sure it spreads globally, in terms of the developing world. So I think there was a very high bar to restrict open source in any way."
www.politico.eu/article/british-deputy-pm-throws-backing-… -
Andrew Ng reposted
Yann LeCun
@ylecun
A nice piece by @AndrewYNg arguing that irrational fears about AI should not cause governments to regulate open source AI models out of existence.
Andrew Ng
@AndrewYNg
My greatest fear for the future of AI is if overhyped risks (such as human extinction) lets tech lobbyists get enacted stifling regulations that suppress open-source and crush innovation.
Read more in our Halloween special issue of the Batch: www.deeplearning.ai/the-batch/issue-220/ -
Andrew Ng
@AndrewYNg
Bill Gurley @bgurley has a great talk on regulatory capture and how it's crushed competition and innovation in many industries. How can we stop this pattern from playing out in AI regulation?
Here's his talk: www.youtube.com/watch?v=F9cO3-MLHOM -
Andrew Ng
@AndrewYNg
My new course, Generative AI for Everyone, is now available!
Learn how Generative AI works, how to use it in professional or personal settings, and how it will affect jobs, businesses and society. This course is accessible to everyone, and assumes no prior coding or AI experience.
Please access it here: www.deeplearning.ai/courses/generative-ai-for-everyone/ -
Andrew Ng
@AndrewYNg
Laws to ensure AI applications are safe, fair, and transparent are needed. But the White House's use of the Defense Production Act—typically reserved for war or national emergencies—distorts AI through the lens of security, for example with phrases like "companies developing any foundation model that poses a serious risk to national security."
Yes, AI -- like many technologies such as electricity and encryption -- is dual use in the sense it can be used for civilian or military purposes. But conflating AI safety for civilian use cases and military applications is a mistake.
It’s also a mistake to set reporting requirements based on a computation threshold for model training. This will stifle open source and innovation: (i) Today’s supercomputer is tomorrow’s pocket watch. So as AI progresses, more players -- including small companies without the compliance capabilities of big tech -- will run into this threshold. (ii) Over time, governments’ reporting requirements tend to become more burdensome. (Ask yourself: Has the tax code become more, or less, complicated over time?)
The right place to regulate AI is at the application layer. Requiring AI applications such as underwriting software, healthcare applications, self-driving, chat applications, etc. to meet stringent requirements, even pass audits, can ensure safety. But adding burdens to foundation model development unnecessarily slows down AI’s progress.
While the White House order isn’t currently stifling startups and open source, it seems to be a step in that direction. The devil will be in the details as its implementation gets fleshed out -- no doubt with assistance from lobbyists -- and I see a lot of risk of missteps. I welcome good regulation to promote responsible AI, and hope the White House can get there. -
Geoffrey Hinton
@geoffreyhinton
Andrew Ng is claiming that the idea that AI could make us extinct is a big-tech conspiracy. A datapoint that does not fit this conspiracy theory is that I left Google so that I could speak freely about the existential threat.
Andrew Ng
@AndrewYNg
I didn't say it is a conspiracy.
But I think overhyped fears about AI leading to human extinction are causing real harm:
- I'm seeing young students discouraged from entering AI because they don't want to contribute to human extinction.
- Hype about harm is also being used to promote bad regulation worldwide, such as requiring licensing of large models, which will crush open-source and stifle innovation.
I know you're sincere in your concerns about AI and human extinction. I just respectfully disagree with you on extinction risk, and also think these arguments -- sincere though they be in your case -- do more harm than good. -
Andrew Ng
@AndrewYNg
My greatest fear for the future of AI is if overhyped risks (such as human extinction) lets tech lobbyists get enacted stifling regulations that suppress open-source and crush innovation.
Read more in our Halloween special issue of the Batch: www.deeplearning.ai/the-batch/issue-220/ -
Andrew Ng
@AndrewYNg
Announcing LangChain CEO @hwchase17’s new short course, Functions, Tools, and Agents with LangChain! Covers latest LLM function calling, how to use LLMs to process formatted data (such as JSON) to integrate with other software, and building agents. www.deeplearning.ai/short-courses/functions-tools-agents-…
-
Andrew Ng
@AndrewYNg
Join me for a livestream, ”Building Computer Vision Applications," on Monday, Nov 6, hosted by Landing AI! Please register here: landing.ai/building-computer-vision-applications/
I’ll discuss the practical aspects of building a vision application, including how to
(i) Identify and scope a vision project,
(ii) Choose the right model and project type, and apply Data-centric AI,
(iii) Deploy an MLOps pipeline. -
Andrew Ng
@AndrewYNg
I’m teaching a new course: Generative AI for Everyone, which is coming soon!
Learn how Generative AI works, how to use it in professional or personal settings, and how it will affect jobs, businesses and society. This course is accessible to everyone, and assumes no prior coding or AI experience. Please join the waitlist to be among the first to know when it goes live!
www.deeplearning.ai/courses/generative-ai-for-everyone -
Andrew Ng
@AndrewYNg
Attending @geoffreyhinton’s retirement celebration at Google with old friends. Thank you for everything you’ve done for AI! @JeffDean @quocleix
-
Andrew Ng
@AndrewYNg
The continued growth of violence since Hamas’ outrageous terrorist attack last Saturday is heartbreaking. I hope everyone will work to protect innocent civilian lives, and build a brighter future for both Israelis and Palestinians.
-
Andrew Ng
@AndrewYNg
While AI can’t solve climate change, it will be a key part of the solution.
I’m thrilled to see the work of Dylan Keil and team at @BearingAI on reducing carbon emissions in shipping. Ships guided by Bearing have sailed over 75 million miles — the equivalent of sailing 3000x around the planet. That’s over a million tons of carbon emissions saved, plus a lot of data to keep training the system!
Bearing AI
@BearingAI
Can generative AI help with climate change?
A new UN report highlights the urgent need to decarbonize shipping, estimating that the investment needed could top $2 trillion (unctad.org/rmt2023).
But Bearing AI is already addressing this today, using AI to optimize existing vessels' operations to reduce fuel and emissions by ~10%.
If Bearing’s software were deployed to the global fleet, it would be equivalent to swapping out 6,000+ vessels for carbon-neutral vessels overnight (the price of one new carbon-neutral vessel can easily be $50+m). -
Andrew Ng
@AndrewYNg
Many AI-savvy programmers are now coding very differently than before, by using LLMs to help with their work. You’ll learn these emerging best practices in “Pair Programming with a Large Language Model,” by Google's @lmoroney.
This short course covers using LLMs to simplify and improve your code, assist with debugging, and minimize technical debt by having AI document your code. The use of LLMs as a programming companion is an important shift that's well worth every developer staying on top of. Please check this out! www.deeplearning.ai/short-courses/pair-programming-llm/ -
Andrew Ng
@AndrewYNg
Looking for AI project ideas for a large company or industry sector? I share in The Batch a brainstorming recipe: (i) break jobs down into tasks, (ii) evaluate common tasks for AI assistance or automation, (iii) assess value. Check out details & examples here: www.deeplearning.ai/the-batch/issue-215/
This is based on a technique that economists use to analyze AI, that's also used by Workhelix (an AI Fund portfolio company co-founded by @erikbryn @amcafee James Milin @danielrock) to help enterprises assess their generative AI opportunities. -
Andrew Ng
@AndrewYNg
-
Andrew Ng
@AndrewYNg
Thanks everyone for all the guesses! Here’s the big reveal. Of the four clips, the one that was my real voice is…
The 4th clip. The others (1-3) were my voice clone created by @Speechlab_ai.
By the way, my parents guessed 3 and 4. Given that even my parents can’t tell, I think this clone is really good!
P.S. Don’t ask me which parent guessed which. I don’t want to get into trouble with them.😬
Andrew Ng
@AndrewYNg
Something fun with voice using generative AI! Here’re four audio clips of either my voice clone (3 clips) or me (1 clip) telling AI jokes. Can you tell which is the real me? Please reply below! Thanks @Speechlab_ai (an @AI_Fund portfolio company) for building the voice clone.
-
Andrew Ng
@AndrewYNg
I recently spoke about Opportunities in AI (link below). Let’s chat more about this! Reply below with your questions and I’ll answer some on Thursday as well. x.com/i/spaces/1lPKqbElVLLGb
Andrew Ng
@AndrewYNg
Here’s the link to the talk: youtu.be/5p248yoa3oE
-
Andrew Ng
@AndrewYNg
Something fun with voice using generative AI! Here’re four audio clips of either my voice clone (3 clips) or me (1 clip) telling AI jokes. Can you tell which is the real me? Please reply below! Thanks @Speechlab_ai (an @AI_Fund portfolio company) for building the voice clone.
-
Andrew Ng
@AndrewYNg
Our new specialization, AI for Good, teaches a framework for using AI when stakes are high and human lives may hang in the balance. @WWRob presents a practical framework for ML in socially important projects (and products of any kind), and illustrates it with real projects in improving climate change, disaster response, and public health.
AI for Good is designed to be useful whether or not you have coding experience. It does include Python code examples that you can execute and interact with to gain deeper insight into different applications.
Please check it out! www.deeplearning.ai/courses/ai-for-good/ -
Post not available here, view on X
Andrew Ng
@AndrewYNg
Thanks! When I first became a professor at Stanford, I decided that rather worrying about how to make myself successful, I would instead focus on trying to make others successful. This turned out to be a great decision, and to this day this is still how I approach work!
-
Andrew Ng
@AndrewYNg
Here's my short interview with Time100 AI. It talks about how, a decade ago, I was thinking about scaling up neural networks as a small step toward AGI!
Even though AGI is still far away, I'm excited every day to continue the AI journey with my teams @AI_Fund, @DeepLearningAI_ and @LandingAI and with all of you!
time.com/collection/time100-ai/6310614/andrew-ng/ -
Andrew Ng
@AndrewYNg
I'm happy to be on the Time AI 100 list of influential people in AI, and thrilled that 8 others from my Stanford group or other teams I led are also named!
Congratulations to my former advisees and team members:
- Sam Altman @sama (as undergrad, interned with me on RL research at Stanford)
- Dario Amodei @Dario_Amodei (worked with me on DeepSpeech and scaling laws at Baidu)
- Alison Darcy @alisonmdarcy (AI+mental health at Stanford)
- Geoff Hinton @geoffreyhinton (member of my team at Google Brain)
- Lila Ibrahim @lilaibrahim (once my COO at Coursera)
- Neal Khosla @nealkhosla (undergrad advisee at Stanford)
- Richard Socher @RichardSocher (PhD advisee with Chris Manning)
- Ilya Sutskever @ilyasut (postdoc at Stanford)
It's thrilling to see the ongoing impact of everyone's work! ❤️ -
Andrew Ng
@AndrewYNg
New short course with @googlecloud, which Nikita Namjoshi and I teach: Understanding and Applying Text Embeddings with Vertex AI! Come learn about text embeddings, a key part of a GenAI developer’s toolkit. You'll learn to apply embeddings to classification, outlier detection, text clustering, and semantic search. We also show how to combine text generation and semantic search to build a question-answering system. Please come join us! www.deeplearning.ai/short-courses/google-cloud-vertex-ai
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
A little known part of the origin story of LLMs: A lot of early Natural Language Processing research was funded by US intelligence agencies that had an interest in machine translation (and also speech recognition). This funding helped develop benchmarks, grow communities and nurture researcher interest in translation research disproportionate to its commercial application, and led to new foundational technologies motivated by translation.
For example, the transformer paper was originally a paper on mapping from one language to another; now we use transformers to map from a prompt to a response. The most widely used tokenizer (BPE) was also popularized in a paper studying how to translate words not seen in the training set.
More on this piece of NLP history in my letter in The Batch! www.deeplearning.ai/the-batch/issue-212/ -
Andrew Ng
@AndrewYNg
How can business leaders, not just coders, build AI business applications with LLMs? Our new short course, created with Microsoft, is out! In How Business Thinkers Can Start Building AI Plugins With Semantic Kernel, taught by @johnmaeda, you'll learn to use Semantic Kernel and build an AI planner that can automatically select the plugins it needs to create multi-step plans with sophisticated logic. Please sign up here: www.deeplearning.ai/short-courses/microsoft-semantic-kernel
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New short course on Fine-tuning LLMs! Many developers are moving beyond only prompting, to also fine-tuning LLMs - that is, taking a pre-trained model and training it further on your own data, which can deliver superior results inexpensively. In this course, @realSharonZhou, CEO of Lamini (disclosure: I’m a minor shareholder) shows you how to recognize when fine-tuning can be help, and how to train an open-source LLM on your own data. I hope you enjoy the course! www.deeplearning.ai/short-courses/finetuning-large-langua…
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
In addition to prompting LLMs, many developers are now also experimenting with fine-tuning. I describe in The Batch how to choose from the growing menu of options for building applications with LLMs: Prompting, few-shot, fine-tuning, pre-training. www.deeplearning.ai/the-batch/issue-210/
-
Andrew Ng
@AndrewYNg
We just released "Large Language Models with Semantic Search”, built with @cohere, and taught by @JayAlammar and @SerranoAcademy. Search is a key part of many applications. Say, you need to retrieve documents or products in response to a user query; how can LLMs help? You’ll learn about (i) Embeddings, to retrieve a collection of documents loosely related to a query, and (ii) LLM assisted re-ranking, to rank them precisely according to relevance. You’ll also go through code showing how to tie all this together to build a complete search system for retrieving relevant Wikipedia articles. Please check it out!
www.deeplearning.ai/short-courses/large-language-models-s…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Do Large Language Models really "understand" the world, or just give the appearance of understanding? Evidence (e.g., Othello-GPT) shows LLMs build models of how the world works, which makes me comfortable saying they do understand. More in The Batch: www.deeplearning.ai/the-batch/issue-209/
-
Andrew Ng
@AndrewYNg
Evaluating and Debugging Generative AI, created with @weights_biases and taught by @carey_phelps, is now available!
ML is an iterative process, and many of us used to track our models/experiments in a text file or spreadsheet, which gets messy. There’re now much better tools like Weights & Biases, that make your development process more efficient. Learn how to use such tools to be systematic how you build Generative AI projects! www.deeplearning.ai/short-courses/evaluating-debugging-ge… -
Andrew Ng
@AndrewYNg
New course with @huggingface! Building Generative AI Applications with Gradio, taught by Apolinário Passos @apolinariosteps, shows you how to quickly create demos of your machine learning applications to test and iterate/share with others. Check it out! www.deeplearning.ai/short-courses/building-generative-ai-…
-
Andrew Ng reposted
LandingAI
@LandingAI
Landing AI’s #sdk makes it easy to integrate your custom #computervision models into your applications.
#machinelearning engineer @DillonLaird walks through a fun example of using computer vision to check if someone’s left a door open.
landing.ai/build-your-custom-computer-vision-app-with-lan… -
Andrew Ng
@AndrewYNg
Looking forward to traveling this and next week to talk about AI! Here’s where I’ll be:
- Japan 🇯🇵 Jul 13-15
- Singapore 🇸🇬 Jul 16-18
- Korea 🇰🇷 Jul 19-21
I’m also looking forward to eating some authentic 親子丼 (Oyakodon), Chilli Crab, and 비빔밥 (Bibimbap)! -
Andrew Ng reposted
Harrison Chase
@hwchase17
🎉New @DeepLearningAI_ class🎉
I had so much fun teaching the last one with @AndrewYNg I had to return for a follow up
This one is a deep dive on the most popular applications of LLMs to date: using them to chat with your data
What do we cover? 👇
www.deeplearning.ai/short-courses/langchain-chat-with-you… -
Andrew Ng
@AndrewYNg
LangChain: Chat with Your Data, a new free short course created with @hwchase17, is now available! www.deeplearning.ai/short-courses/langchain-chat-with-you…
In this 1 hour course, you’ll learn how to build one of the most requested LLM-based applications: Answering questions using information from a document or collection of documents (often called Retrieval Augmented Generation). You'll also learn how to use vector stores and embeddings to retrieve document chunks relevant to a query. I hope you enjoy the course!This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
New course! Generative AI with Large Language Models, created with @AWS and hosted on @coursera. This course goes deep into the technical foundations of LLMs and how to use them. You can sign up here: deeplearning.ai/courses/generative-ai-with-llms
You’ll work through the full life-cycle of a generative AI project, and learn specific techniques like RLHF; zero-shot, one-shot, and few-shot learning with LLMs; advanced prompting frameworks like ReAct; even fine-tuning LLMs, and gain hands-on practice with all of these techniques.
Instructors @anbarth @cfregly @Shelbee_SE and @mikegchambers all do incredible Generative AI work at AWS, and have supported many companies to build creative LLM applications. They bring tremendous practical LLM expertise to this course.
I'm confident you’ll finish this course with a deeper understanding of how LLMs work, and how to use them. I hope you enjoy the course! -
Andrew Ng
@AndrewYNg
Hello Vancouver! Looking to speaking at the CVPR conference's Computer Vision in the Wild workshop tomorrow morning, on Visual Prompting! computer-vision-in-the-wild.github.io/cvpr-2023/
Jianwei Yang
@jw2yang4ai
We are super excited to announce Dr. @AndrewYNg will give a keynote talk at our 2nd CVinW Workshop! Dr. Ng is the pioneer of deep learning, founder of Coursera&LandingAI. His recent focus on data-centric AI for vision is perfectly aligned with our workshop! Join us on June 19th!
-
Andrew Ng
@AndrewYNg
AI art isn't ready yet to take over the world! My 4 year old daughter made this lovely camera -- with removable polaroid picture action -- for me for Father's Day, and I love this so much more than anything Midjourney's ever generated. ❤️
Happy Father's Day everyone! -
Andrew Ng
@AndrewYNg
AI and coding is now so important to society, and gaining AI and coding skills is so empowering of individuals, that schools should teach it to every child. My piece with @Andy_Pasinetti on this. venturebeat.com/ai/schools-should-teach-ai-to-every-child…
-
Andrew Ng reposted
Geoffrey Hinton
@geoffreyhinton
I used to talk to Andrew a lot and it was great to catch up again and get his take on the various risks posed by recent developments in AI. We agreed on a lot of things, especially on the need for the researchers to arrive at a consensus view of the risks to inform policy makers.
-
Andrew Ng
@AndrewYNg
Had an insightful conversation with @geoffreyhinton about AI and catastrophic risks. Two thoughts we want to share:
(i) It's important that AI scientists reach consensus on risks-similar to climate scientists, who have rough consensus on climate change-to shape good policy.
(ii) Do AI models understand the world? We think they do. If we list out and develop a shared view on key technical questions like this, it will help move us toward consensus on risks.
I learned a lot speaking with Geoff. Let’s all of us in AI keep having conversations to learn from each other!This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Had a great conversation with Yoshua Bengio. Both of us agreed that a good step forward for AI risk is to articulate the concrete scenarios where AI can lead to significant harm. More to come, and looking forward to continuing the conversation!
-
Andrew Ng
@AndrewYNg
I'd like to have a real conversation about whether AI is a risk for human extinction. Honestly, I don't get how AI poses this risk.
What are your thoughts? And, who do you think has a thoughtful perspective on how AI poses this risk that I should talk to? -
Andrew Ng
@AndrewYNg
1/The false and sensationalist coverage of the purported Air Force simulation where an AI-drone decided to kill an operator will be remembered as a highly regrettable episode of AI doomsaying hype. Lets be honest about what are, and what are not, real risks.
Andrew Ng
@AndrewYNg
2/Developers who’re shipping AI products do see real risks like bias, fairness, inaccuracies, job displacement, and are working to address them. Unrealistic hype distracts from the real problems. It also discourages people from entering AI and building things that help people.
Andrew Ng
@AndrewYNg
3/Here’s a Vice article that describes in more detail what happened. www.vice.com/en/article/4a33gj/ai-controlled-drone-goes-r…
-
Andrew Ng reposted
Isa Fulford
@isafulf
You’ll build a customer service assistant that can look up relevant information from a product catalogue!
DeepLearning.AI
@DeepLearningAI
Building Systems with the ChatGPT API is live!
In this short course, you’ll learn how to break a complex task down to be carried out via multiple API calls to an LLM.
Join for free: www.deeplearning.ai/courses/chatgpt-building-system?hss_c… -
Andrew Ng reposted
Sharon Zhou
@realSharonZhou
Launched 🚀 a short-course on diffusion models, with my good friend @AndrewYNg!
✅ Get the foundational intuition
✅ Run the code
✅ Literally all in < 1 hr!
👕 Bonus points: I tried to match Andrew's classic blue shirt, you should too
www.deeplearning.ai/short-courses/how-diffusion-models-work/
Andrew Ng
@AndrewYNg
1/Thrilled to announce: 3 new Generative AI courses!
* Building Systems with the ChatGPT API, with OpenAI’s @isafulf
* LangChain for LLM Application Development, with LangChain’s @hwchase17
* How Diffusion Models Work, by @realSharonZhou
Check them out: deeplearning.ai/short-courses/ -
Andrew Ng
@AndrewYNg
1/Thrilled to announce: 3 new Generative AI courses!
* Building Systems with the ChatGPT API, with OpenAI’s @isafulf
* LangChain for LLM Application Development, with LangChain’s @hwchase17
* How Diffusion Models Work, by @realSharonZhou
Check them out: deeplearning.ai/short-courses/Post not available here, view on X
Andrew Ng
@AndrewYNg
4/How Diffusion Models Work: Learn the technical details of how diffusion models - which power Midjourney, DALL·E 2, and Stable Diffusion - work. You’ll also end up with working code to generate your own video game sprites in Jupyter! learn.deeplearning.ai/diffusion-models
Andrew Ng
@AndrewYNg
5/All of these courses are free for a limited time, and each can be completed in around 1-1.5 hours. I hope you enjoy them. Also, please help me spread the word and encourage others to sign up too!
-
Andrew Ng
@AndrewYNg
When I think of existential risks to large parts of humanity:
* The next pandemic
* Climate change→massive depopulation
* Another asteroid
AI will be a key part of our solution. So if you want humanity to survive & thrive the next 1000 years, lets make AI go faster, not slower. -
Andrew Ng
@AndrewYNg
Thank you NASDAQ Tradetalks for having me to talk about @AI_Fund. AI is a general purpose technology that's made new business use cases possible in many sectors: maritime shipping, K12 education, real estate, .... AI fund is building one company a month with great entrepreneurs to pursue these exciting opportunities!
TradeTalks
@TradeTalks
AI Fund Managing General Partner @AndrewYNg joins @JillMalandrino on @Nasdaq #tradetalks from @SALTConference @iConnections_io to discuss the biggest opportunities in the #ai startups landscape. x.com/i/broadcasts/1ZkKzXrjkDvJv
-
Andrew Ng reposted
Stanford Online
@StanfordOnline
Recently Prof. Fei-Fei Li sat down with @AndrewYNg of @DeepLearningAI_ to discuss the past and present state of human-centered #ai, and some of the "audacious questions" yet to be answered. Full interview 👉 youtu.be/UNhC6Ox0T0o @drfeifei @coursera @StanfordHAI @StanfordAILab
-
Andrew Ng
@AndrewYNg
"Give me your tired, your poor, Your huddled masses yearning to breathe free," reads a poem on the Statue of Liberty. 🗽
Looking at the immigration chaos on the southern US border, I wish we can build a compassionate world that supports and lifts up anyone in need. -
Andrew Ng
@AndrewYNg
Over 200,000 people have signed up for our new short course! ChatGPT prompt engineering for developers, taught with @OpenAI @isafulf.
You can still access it for free here: www.deeplearning.ai/short-courses/chatgpt-prompt-engineer… -
Andrew Ng
@AndrewYNg
It was a privilege working with you @geoffreyhinton in the early days of Google Brain. Thank you for your immeasurable contributions both to AI technology and for being a sensible voice on the risks of AI, such as misinformation, autonomous weapons and labor disruption.
Geoffrey Hinton
@geoffreyhinton
In the NYT today, Cade Metz implies that I left Google so that I could criticize Google. Actually, I left so that I could talk about the dangers of AI without considering how this impacts Google. Google has acted very responsibly.
-
Andrew Ng
@AndrewYNg
1/ Thrilled to announce: Our new course ChatGPT Prompt Engineering for Developers, created together with @OpenAI, is available now for free! Access it here: www.deeplearning.ai/short-courses/chatgpt-prompt-engineer…
This media couldn't be displayed. See it on the original post.
Andrew Ng
@AndrewYNg
2/ In this short, 1.5 hour course, @isafulf and I teach how you can build applications quickly using large language models (LLMs), and best practices for doing so. Thank you Isa and the @OpenAI team for collaborating on this exciting course.
Andrew Ng
@AndrewYNg
3/ Until now, there has been relatively little material on best practices for developers wanting to build applications using API access to LLMs (as opposed to how to prompt the web UI). We’re excited to change that. I hope you enjoy the course!
-
Andrew Ng reposted
Joselo
@joselobenitezg
This is amazing! We are undoubtedly experiencing the “GPT-3 moment” in the field of Computer Vision 🔥🚀
Andrew Ng
@AndrewYNg
1/2 Visual Prompting takes ideas from text prompting, which has transformed NLP, to computer vision. Via a simple "visual prompt", you can build a vision model in seconds. Try it out here: app.landing.ai/public/visual_prompting
-
Andrew Ng
@AndrewYNg
1/2 Visual Prompting takes ideas from text prompting, which has transformed NLP, to computer vision. Via a simple "visual prompt", you can build a vision model in seconds. Try it out here: app.landing.ai/public/visual_prompting
Andrew Ng
@AndrewYNg
2/2 Also, here's a video of the livestream where I discuss video prompting in detail, including why I think computer vision is reaching an inflection point (because of large pre-trained vision transformers). Let me know what you think! www.youtube.com/watch?v=FE88OOUBonQ
-
Andrew Ng
@AndrewYNg
Join me Tuesday to discuss Visual Prompting! Text prompting has revolutionized NLP:
Traditional AI: Label -> Train -> Predict
Prompt based AI: Prompt -> Predict
I’ll share early thoughts on taking this from text to vision w/Visual Prompting. RSVP here: landing.ai/visual-prompting -
Andrew Ng
@AndrewYNg
1/Many AI businesses are searching for moats. A data moat strengthens your defenses if the data is hard to replicate. But if a business offers direct access to a ML model's output (like many LLMs), competitors can use that output as labeled training data, circumventing the moat.
Andrew Ng
@AndrewYNg
2/While a data moat can be helpful, I find people tend to overestimate its strength. The engineering recipe of training on someone else’s API output also raises technical, business and legal questions. My longer piece on this in The Batch: www.deeplearning.ai/the-batch/issue-200/
-
Andrew Ng reposted
Yann LeCun
@ylecun
The gist of our arguments against the 6-month AI moratorium at @VentureBeat.
With @AndrewYNg.
venturebeat.com/ai/titans-of-ai-industry-andrew-ng-and-ya… -
Andrew Ng
@AndrewYNg
Here's a video of my conversation with @ylecun about the proposed 6 month AI pause. I would love to hear what you think of the AI pause -- please reply to share your thoughts! www.youtube.com/watch?v=BY9KV8uCtj4
-
Andrew Ng
@AndrewYNg
Yann LeCun and I have thought a lot about the proposed 6 month AI pause, and plan to chat about it on Friday - the questions it raises for AI and its impact on developers and companies. Please join @ylecun and me for the conversation! RSVP here: 6monthaipause.eventbrite.com
-
Andrew Ng
@AndrewYNg
Thanks @ClementDelangue and the @huggingface team for hosting the "Woodstock of AI" in San Francisco last night. Great event, and it was wonderful to see all the enthusiasm for AI and the community's rapid technical & product experimentation!
Brigitte 🤗
@BrigitteTousi
Great seeing @AndrewYNg at our OS AI meetup! @ClementDelangue @huggingface 🤗 #woodstockai
-
Andrew Ng
@AndrewYNg
1/The call for a 6 month moratorium on making AI progress beyond GPT-4 is a terrible idea.
I'm seeing many new applications in education, healthcare, food, ... that'll help many people. Improving GPT-4 will help. Lets balance the huge value AI is creating vs. realistic risks.Post not available here, view on X
Andrew Ng
@AndrewYNg
3/Responsible AI is important, and AI has risks. The popular press narrative that AI companies are running amok shipping unsafe code is just not true. The vast majority (sadly, not all) of AI teams take responsible AI and safety seriously.
Andrew Ng
@AndrewYNg
4/A 6 month moratorium is not a practical proposal. To advance AI safety, regulations around transparency and auditing would be more practical and make a bigger difference. Let's also invest more in safety while we advance the technology, rather than stifle progress.
-
Andrew Ng
@AndrewYNg
Workera uses generative AI to develop assessments to measure technical skills. (Assessments are vetted by expert humans before being used.) This helps companies and individuals get a granular skill assessment, to decide where to invest in development and grow careers more efficiently.
Congratulations to my friend @kiankatan and @workera_ (an @AI_Fund portfolio company) for raising its $23.5M Series B to keep growing its skills intelligence platform! techcrunch.com/2023/03/23/upskilling-platform-workera-rai… -
Andrew Ng
@AndrewYNg
If you’re a frequent user of ChatGPT, what are you using it for?
-
Andrew Ng
@AndrewYNg
Dealing with SVB's collapse, I'm seeing a lot of the best of the AI & startup world: People are helping each other, sharing tips, even wiring funds on trust and without docs so someone else can make payroll. Very encouraging seeing our community rise together to the challenge!
-
Andrew Ng
@AndrewYNg
I'm glad US gov is looking into protecting SVB's depositors. I see no need to bail out the bank's investors, but failing to protect depositors - who had taken no reckless risks - can mean large layoffs, damage to trust in banks, and setting back AI & tech innovation by years.
-
Andrew Ng
@AndrewYNg
Sending love and support to everyone at SVB as well as everyone worried about their funds. I know today will be a tough day for a lot of people. ❤️
-
Andrew Ng
@AndrewYNg
One year from now, what size LLM (large language model) do you think will be used for the most inferences?
-
Andrew Ng
@AndrewYNg
Building and deploying a machine learning model usually takes months. How can you go from starting a project to training and deploying your model in minutes? Here's a 3min overview of the LandingLens platform.
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Good point François!
We can make exciting progress with deep learning without hyping things up via tenuous connections to AGI.
François Chollet
@fchollet
When companies that train deep learning models talk about AGI, it's as if a 3D printing company talked about how the next generation of the technology was going to bring universal abundance by enabling arbitrary matter replication -- if we can avoid the grey goo scenario
-
Andrew Ng
@AndrewYNg
Cool WSJ piece on how tools like Landing AI's computer vision platform make it easier to build AI systems. @IsabelleBiscuit www.wsj.com/articles/pressure-mounts-on-cios-to-build-mor…
Andrew Ng
@AndrewYNg
If you're wondering what's happening in the picture, I was doing a LandingLens demo on fire detection using a real fire. 🔥
As one would say in Gen Z slang, AI is Fire! 😀 -
Andrew Ng
@AndrewYNg
1/Landing AI is releasing our computer vision tool for anyone to use! You can label images, train a model, and deploy your model to production in minutes. Our data-centric tool then helps you improve your model quickly. Start for free at landing.ai
Andrew Ng
@AndrewYNg
2/We’re on a mission to democratize access to AI creation, so everyone can easily train and deploy custom ML models on their own data. Every business and every person now has their own data, so using only models that someone else had trained on their data is not good enough.
Andrew Ng
@AndrewYNg
3/But building and deploying ML systems is hard. After years of work, I’m thrilled to finally offer LandingLens to everyone. Computer vision projects that used to take me a year are now done in days/weeks on LandingLens. Please try it out and also help me spread the word!
-
Andrew Ng
@AndrewYNg
Just as Instagram’s parade of perfect physiques makes people feel they don’t measure up, AI’s parade of exciting projects can make people feel their own projects are lacking. Don’t let shiny objects make you doubt the worth of your work!
Read more: www.deeplearning.ai/the-batch/issue-184/ -
Andrew Ng
@AndrewYNg
Happy Valentine's Day -- love you all! ❤️❤️❤️
-
Andrew Ng
@AndrewYNg
With AI surging, business leaders need to guide organizations through rapidly emerging opportunities and threats. I’m delighted FourthBrain is offering a live workshop on this, that I'm looking forward to speaking at.
fourthbrain.ai/ai-business-leaders/ -
Andrew Ng
@AndrewYNg
I wish schools could make homework so joyful that students want to do it themselves, rather than let ChatGPT have all the fun.
-
Andrew Ng
@AndrewYNg
If you're starting a reading group on Large Language Models (LLMs), what is one research paper you will want added to the reading list?
Researchers: Feel free to recommend your own paper too! -
Andrew Ng reposted
Jaspreet Bindra : Homo Promptus
@j_bindra
A good article on whether #llm s like #gpt3 or #chatgpt will replace #search - by the great @AndrewYNg
Short answer (my belief too): Yes, technically possible, but incumbent business model a problem. Classical Innovators Dilemma
www.deeplearning.ai/the-batch/issue-180/
@OpenAI @Google @bing -
Andrew Ng
@AndrewYNg
Which type of Generative AI do you think will have a bigger commercial impact: Generating images (e.g., diffusion algorithms, stable diffusion) or text (e.g., LLMs, ChatGPT)?
-
Andrew Ng
@AndrewYNg
1/Math for Machine Learning and Data Science is now available on Coursera! Taught by @luis_likes_math, this gives an intuitive understanding of the most important math concepts for AI.
www.coursera.org/specializations/mathematics-for-machine-…
Andrew Ng
@AndrewYNg
2/I’ve often said “don’t worry about it” when it comes to math, because math shouldn’t hold anyone back from making progress in ML. And, understanding some key topics in linear algebra, calculus, and prob & stats will help you better get learning algorithms to work.
Andrew Ng
@AndrewYNg
3/This specialization was designed with numerous interactive visualizations to help you see how the math works. Math isn’t about memorizing formulas; it’s about sharpening your intuition. I hope you enjoy the specialization!
-
Andrew Ng
@AndrewYNg
Do large language models like ChatGPT have a shot at disrupting search engines? I share my take here. (Short answer: Yes. But the details are complicated.) www.deeplearning.ai/the-batch/issue-180/
-
Andrew Ng
@AndrewYNg
1/2 Thanks everyone that replied to this. Lots of great ideas here! The diversity of our hopes for AI is a sign of how large our field is -- we collectively want AI to go lots of places.
Andrew Ng
@AndrewYNg
What do you hope for in AI in 2023?
Andrew Ng
@AndrewYNg
2/2 I also wrote a summary in The Batch of the common themes in the hopes people shared. Check it out here. www.deeplearning.ai/the-batch/issue-178/
-
Andrew Ng
@AndrewYNg
1/Several leaders share their hopes for AI in 2023, including finding key missing pieces that will enable algorithms to reason, building a personal data timeline, improving AI processes, discovering new principles for explainability, and using generative AI for active learning.
Post not available here, view on X
Andrew Ng
@AndrewYNg
6/Reza Zadeh sees generative AI bringing progress to active learning, where a system picks its own examples to be labeled to improve the data. With generative AI, he sees a potential revolution in algorithms generating new data to request to be labeled.
www.deeplearning.ai/the-batch/reza-zadeh-active-learning-…
Andrew Ng
@AndrewYNg
7/Exciting times ahead, and thank you Yoshua Bengio, @AlonHalevy, @douwekiela, @_beenkim, and @Reza_Zadeh for writing for The Batch!
-
Andrew Ng
@AndrewYNg
What do you hope for in AI in 2023?
-
Andrew Ng
@AndrewYNg
2022 has been a great year for AI progress. Excitingly, this has laid a foundation for even more progress in 2023. Happy New Year, friends! 🎉
-
Andrew Ng reposted
iyànú
@iyanugabriel__
"How to build your career in AI" by Andrew Ng is a must-read for anyone interested in starting/growing one's career in Artificial Intelligence.
I particularly love chapters 7 & 11 - "A Simple Framework for Starting Your AI Job Search" and "On Overcoming Imposter Syndrome" -
Andrew Ng
@AndrewYNg
Despite the cold weather, we’re in the middle of a red hot AI summer.
Read my thoughts — and a roundup of this year’s biggest AI news — in The Batch! www.deeplearning.ai/the-batch/issue-176/ -
Andrew Ng
@AndrewYNg
1/I find AI’s role in the FIFA World Cup has made the game more fun! ⚽️
Post not available here, view on X
Andrew Ng
@AndrewYNg
5/For the World Cup final, here’s wishing good luck to the Argentina team 🇦🇷, The France team 🇫🇷, and the AI engineering team 🤖!
Andrew Ng
@AndrewYNg
6/More about this in The Batch. www.deeplearning.ai/the-batch/the-world-cups-ai-referee/
-
Andrew Ng
@AndrewYNg
New ebook! How can you build a career in AI?
This compiles my tips on how to:
∙ Learn technical skills 🎓
∙ Choose and build AI projects 🧑💻
∙ Find a job 🦾
Let me know your feedback!
info.deeplearning.ai/how-to-build-a-career-in-ai-book -
Andrew Ng
@AndrewYNg
New course announcement! Math for Machine Learning and Data Science.
Designed to make learning AI's math easy:
∙ Linear algebra
∙ Probability & Stats
∙ Calculus
Will help learners build better ML systems and pass job interviews!
www.deeplearning.ai/courses/mathematics-for-machine-learn… -
Andrew Ng
@AndrewYNg
1/Large language models like Galactica and ChatGPT can spout nonsense in a confident, authoritative tone. This overconfidence - which reflects the data they’re trained on - makes them more likely to mislead.
Post not available here, view on X
Andrew Ng
@AndrewYNg
3/Building large language models that can accurately decide when to be confident and when not to will reduce their risk of misinformation and build trust.
Andrew Ng
@AndrewYNg
4/More on this in The Batch. www.deeplearning.ai/the-batch/issue-174/
-
Andrew Ng
@AndrewYNg
I feel lucky to be alive when technological marvels are invented every week. The rapid progress makes life more fun!
-
Andrew Ng
@AndrewYNg
ChatGPT is sometimes amazing, and sometimes hilariously wrong. Here, it confidently explains to me why an abacus is faster than a GPU. 😃
-
Andrew Ng
@AndrewYNg
Congratulations Mark Chrystal and the @NetailAI team on closing $5M in funding to build AI to help retailers with pricing analytics. This is an exciting application of machine learning and will help many businesses. @AI_Fund @magaracvc www.bizjournals.com/pittsburgh/inno/stories/Inno-insights…
-
Andrew Ng
@AndrewYNg
AI twitter had erupted in controversy over Meta’s release of the Galactica demo. Is it a dangerous product, or was it unnecessarily maligned with overblown fears of harm? Here’s my take. www.deeplearning.ai/the-batch/issue-172/
-
Andrew Ng
@AndrewYNg
A new UN Climate Change report estimates we might be on track to 2.5°C of warming. It's time to seriously consider cooling down Earth by geoengineering using stratospheric aerosol injection, a solution in which AI will play a critical role. See why. www.deeplearning.ai/the-batch/issue-169/
-
Andrew Ng
@AndrewYNg
Halloween is coming! 🎃 Check out our special issue of The Batch on this year's biggest AI fears. 👻www.deeplearning.ai/the-batch/issue-168/
-
Andrew Ng
@AndrewYNg
I’d like to address the serious matter of some newcomers to AI experiencing imposter syndrome, where someone wonders if they’re a fraud or really belong in the AI community. Lets build a community that encourages and welcomes everyone. www.deeplearning.ai/the-batch/issue-164/
-
Andrew Ng reposted
Greg Lavender
@GregL_Intel
I am looking forward to having @AndrewYNg, founder of Landing AI, join us at Intel Innovation next week as our luminary speaker. Hear the important role Data-Centric AI plays in increasing openness in AI. Learn more and sign up at: intel.ly/3RWWm8Y #intelon
-
Andrew Ng
@AndrewYNg
The open way that Stable Diffusion's image generation model was released -- allowing users to it on their own machines, not just via API -- has made it a landmark event for AI. Kudos to @EMostaque and the @StableDiffusion team! More in The Batch. read.deeplearning.ai/the-batch/issue-162/
-
Andrew Ng
@AndrewYNg
Today is Labor Day in the U.S. I wish for all workers everywhere that we will all find work that is meaningful and joyful.
-
Andrew Ng
@AndrewYNg
Finding an AI job can be difficult. By paying close attention to certain details during your search, you can raise your odds of finding a role, team, and company that you truly love. Read more in #thebatch. read.deeplearning.ai/the-batch/issue-160/
-
Andrew Ng
@AndrewYNg
Interested in an AI job, but unsure what different roles entail? Informal chats with working professionals can help you narrow your job search. Read my tips on informational interviews (which are different from job interviews) in The Batch. read.deeplearning.ai/the-batch/issue-159/
-
Andrew Ng
@AndrewYNg
1/2 Biden's $10,000 student debt forgiveness announcement is a reminder we need to make the education system better. Education needs to become more affordable, more inclusive, and less exclusionary. We also need to pay more attention to helping students find good jobs.
Andrew Ng
@AndrewYNg
2/2 Good education really changes lives, but we also can’t continue to have students take out loans to pay for programs that don’t deliver hoped-for benefits. The current system isn’t sustainable, and we can and must do better.
-
Andrew Ng
@AndrewYNg
Here's a framework useful for searching for an AI job. Especially for people entering the field, consider if you’re switching roles and/or industries. Please take a look, and if you have any other AI job search tips to share, please reply to this tweet! read.deeplearning.ai/the-batch/build-career-part-5/
-
Andrew Ng
@AndrewYNg
Happy 75th Independence Day to all my friends in India! 🇮🇳Democracy is a powerful tool for improving people's lives. It is wonderful to see the many people in India and worldwide continuing to fight for this ideal.
-
Andrew Ng
@AndrewYNg
Reinforcement learning (RL) algorithms are quite finicky -- sensitive to picking hard-to-tune hyperparameters -- compared to supervised deep learning, which has become much more robust the past decade. Will RL progress similarly? My thoughts in The Batch: read.deeplearning.ai/the-batch/issue-156/
-
Andrew Ng
@AndrewYNg
Course 3 of the new Machine Learning Specialization is now available on @Coursera! We'd launched Courses 1&2 last month, and this final course teaches unsupervised learning, recommender systems, and reinforcement learning. Check it out here: www.coursera.org/specializations/machine-learning-introdu…
-
Lora Aroyo
@laroyo
Joint effort across multiple research organizations to build the #datacentricai --> @AndrewYNg presenting an #invitedtalk @icmlconf #dataperf #workshop sites.google.com/corp/view/dataperf2022/home @DataCentricAI
Andrew Ng
@AndrewYNg
Thanks @laroyo and team for having me speak at the ICML DataPerf workshop. A lot of people had contributed to the Data-Centric AI competition last year, and I was glad to be able to share the story behind the competition.
-
Andrew Ng
@AndrewYNg
To advance a career in AI, how should one pick projects to work on? Your career will include a sequence of projects growing in scope and complexity. So it helps to pick what can be a stepping stone. My tips on this career progression process in The Batch. read.deeplearning.ai/the-batch/how-to-build-a-career-in-a…
-
Andrew Ng
@AndrewYNg
Interested in accelerating your AI career? Join @alliekmiller, @coursera’s Shravan Goli and me on 8/3 for an interactive discussion about skills that will help you succeed. Reserve your spot here: youraicareer.eventbrite.com
-
Andrew Ng
@AndrewYNg
How do you pick projects to advance in a technical career in AI? Your career will include many projects, hopefully growing in scope, complexity, and impact over time. Picking good projects will create stepping stones for you to move forward and up. Tips in this week's The Batch.
Andrew Ng
@AndrewYNg
Here's the link: read.deeplearning.ai/the-batch/issue-153/
-
Andrew Ng
@AndrewYNg
What are the most important topics to study for building a technical career in AI? I share my thoughts on this in The Batch. read.deeplearning.ai/the-batch/issue-152/
-
Andrew Ng
@AndrewYNg
Thanks! Last week I wrote about key steps to building a technical career in AI. The full letter is here: read.deeplearning.ai/the-batch/issue-151/
Will have more to say on this in this week's The Batch!
𝗩𝗜𝗖𝗞𝕏
@VictorOmondiCDS
Loving this pic from The Batch (@AndrewYNg ) Deep learning AI newsletter showing key steps to career growth.
* Foundation: Community
* Pillars: Learning, projects, job search. -
Andrew Ng
@AndrewYNg
The new Machine Learning Specialization by @DeepLearningAI_ & @StanfordOnline is now available on @Coursera!
You’ll learn the fundamentals of ML & gain practical experience with Python. Please help me spread the word, and encourage others to take ML! www.coursera.org/specializations/machine-learning-introdu…This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Kira Learning is building the go-to solution for K-12 educators and administrators looking for standards aligned Computer Science curriculum and learning tools. @DeepLearningAI_ & @AIFund are proud to support @Kira_Learning. Kudos to @Andy_Pasinetti and team!
Kira
@Kira_Learning
Today we’re excited to announce a $6M Seed Round led by @NEA & @AndrewyNg’s @AI_Fund!
Check out our blog for more on today's news, and stay tuned for more exciting announcements to come!
kira-learning.com/blogDetail/?id=28 -
Andrew Ng reposted
Rick Merritt
@rickbmerritt
Kudos to @landingAI for taking #computervision to the factory floor, making manufacturing more efficient with data-centric #ai #automation @AndrewYNg blogs.nvidia.com/blog/2022/06/06/andrew-ng-ai-factory-lan…
-
Andrew Ng
@AndrewYNg
How will rising interest rates affect AI? Projects that deliver financial returns within a few years will look more attractive than long-term bets. AI teams should be ready for increased pressure to demonstrate near-term tangible value. More here: read.deeplearning.ai/the-batch/issue-147/
-
Andrew Ng
@AndrewYNg
Read about the surprising origins of many of Machine Learning’s foundational algorithms here. Many algorithms were invented long before digital computers. I also found the first application of logistic regression, K-means and other algorithms fascinating!
read.deeplearning.ai/the-batch/issue-146/ -
Andrew Ng
@AndrewYNg
The call to arm teachers following the Texas school shooting tragedy is an awful idea.
Teachers need to focus on helping students learn, not on preparing for a shootout. The U.S. needs better gun safety laws to protect children.
www.msnbc.com/rachel-maddow-show/maddowblog/school-shooti… -
Andrew Ng
@AndrewYNg
Speaking as someone who has worked on anti-spam against determined adversaries, I found this a nice, thoughtful thread on the reality of spam. Thank you @paraga.
Parag Agrawal
@paraga
Let’s talk about spam. And let’s do so with the benefit of data, facts, and context…
-
Andrew Ng
@AndrewYNg
A few people asked if Coursera and I share a birthday. Yes we do -- April 18th. 😀
Andrew Ng
@AndrewYNg
Coursera had started with my machine learning class. I’m delighted to announce that I’ll soon be teaching a new Machine Learning Specialization, created by @StanfordOnline & @DeepLearningAI_. Please help me spread the word! You can sign up here: www.deeplearning.ai/program/machine-learning-specialization/
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Coursera had started with my machine learning class. I’m delighted to announce that I’ll soon be teaching a new Machine Learning Specialization, created by @StanfordOnline & @DeepLearningAI_. Please help me spread the word! You can sign up here: www.deeplearning.ai/program/machine-learning-specialization/
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Hello New York! Looking forward to speaking at
@insightpartners' ScaleUp conference today.
Also, thanks, um, for putting my face on a giant building. 😀 -
Andrew Ng
@AndrewYNg
Thank you NVIDIA team - it was fun speaking at GTC and sharing the stage with @ctnzr.
Watch the talk where I showed the first public demo of some data-centric features of LandingLens, an MLOps platform for computer vision built by @LandingAI: nvda.ws/3JxcmKH
NVIDIA AI Developer
@NVIDIAAIDev
Today's the day. Starting at 9am PDT: @AndrewYNg will discuss the data-centric #ai movement followed by a live Q&A. Don't miss it. #gtc22 nvda.ws/3JxcmKH
-
Andrew Ng
@AndrewYNg
1/2 I’d started speaking about data-centric AI development (in this youtube video) exactly a year ago - on Mar 24, 2021. It’s great to see the momentum the movement has. www.youtube.com/watch?v=06-AZXmwHjo
Andrew Ng
@AndrewYNg
2/2 Huge thanks to all the developers, researchers, students, entrepreneurs, investors, business leaders, workshop organizers, reporters and others pushing data-centric AI forward!
-
Andrew Ng
@AndrewYNg
I stand with Ukraine 🇺🇦 against Putin's unprovoked invasion.
-
Andrew Ng
@AndrewYNg
To help anyone apply Data-centric AI, I’m excited to announce a new resource hub, with articles by experts (including @AnimaAnandkumar, @msbernst and D Sculley) and talks from the NeurIPS Data-centric AI workshop. Check it out! datacentricai.org
-
Andrew Ng
@AndrewYNg
Thanks everyone for your well wishes last week after I tested positive for Covid. You really made me feel better!
I'm now almost symptom free, and it looks like the virus is clearing out of my system.
Also, Happy Valentine's Day! ❤️ -
Andrew Ng
@AndrewYNg
Thanks @newsbeagle for writing up our interview on Data-Centric AI in @IEEESpectrum. Worth a read for anyone interested in data-centric development, & also touches on foundation models, big data vs. small data, and practical data engineering. spectrum.ieee.org/andrew-ng-data-centric-ai
-
Andrew Ng
@AndrewYNg
I just tested positive for Covid.
I am very grateful to the people that’ve worked on vaccines worldwide. My symptoms are comparable to a mild flu, but if I hadn’t had a vaccine & booster, I’m sure it would have been much worse. Thank you! ❤️ -
Andrew Ng
@AndrewYNg
Happy New Year! Wishing you a healthy, happy and prosperous Year of the Tiger! 🐯
-
Andrew Ng
@AndrewYNg
How can you build a career in AI? This week's special issue of The Batch features guest articles that share advice on this. Please take a look! (1/2) read.deeplearning.ai/the-batch/issue-128/
Andrew Ng
@AndrewYNg
The articles cover:
- How to learn machine learning (by Kian Katanforoosh)
- How to gain practical experience (Julien Despois)
- How to overcome societal obstacles (Benjamin Harvey)
- How to get a job in AI (Luis Serrano)
- How to keep up in a changing field (Eugene Yan)
(2/2) -
Andrew Ng
@AndrewYNg
The @DeepLearningAI_ mentors play a huge role in helping learners. I’d like to invite you to “How to #beadeeplearner” on Jan 26, where a panel of mentors will answer your questions about the Deep Learning Specialization or about building an AI career. beadeeplearner.eventbrite.com
-
Andrew Ng
@AndrewYNg
Given the impact of Computer Science and AI on every person, it is important to introduce K-12 learners to these subjects. I’m thrilled my friend @Andy_Pasinetti (one of the best education leaders I know) is launching @Kira_Learning to transform K-12 computer science!
Andrea Pasinetti
@Andy_Pasinetti
I'm excited to share what I've been working on as a Founder-in-Residence @AI_Fund for the past several months: @Kira_Learning is using AI to reimagine K-12 education publishing, beginning with computer science.
-
Andrew Ng
@AndrewYNg
The PyTorch team has done a great job building a powerful programming framework. Thank you @soumithchintala and everyone on the team for your contributions to deep learning!
Soumith Chintala
@soumithchintala
It’s been 5 years since we launched @pytorch. It’s much bigger than we expected -- usage, contributors, funding. We’re blessed with success, but not perfect. A thread (mirrored at soumith.ch/posts/2022/01/pytorch-retro/ ) about some of the interesting decisions and pivots we’ve had to make 👇
-
Andrew Ng
@AndrewYNg
I'm thrilled to welcome manufacturing machine vision and industrial automation guru @DavidDechow to Landing AI. He's trained up generations of machine vision engineers. Looking forward to his help scaling up deep learning in manufacturing! www.zdnet.com/article/landing-ai-hires-vision-expert-dech…
-
Andrew Ng
@AndrewYNg
Over 1 million learners have taken the Deep Learning Specialization by @DeepLearningAI_ on @Coursera. This month, I’m excited that we're celebrating our learners. I hope their stories will inspire you to continue on your own learning path! www.deeplearning.ai/program/deep-learning-specialization/…
-
Andrew Ng
@AndrewYNg
Today is the one year anniversary of the insurrection at the U.S. Capitol. We cannot take democracy in any country for granted, but must instead keep cherishing and defending it.
-
Andrew Ng
@AndrewYNg
I’m looking forward to 2022. Let’s all of us together make it a great year!
-
Andrew Ng
@AndrewYNg
I’m happy to see MLCommons release both a multilingual speech dataset and a large 30,000 hour diverse english dataset. We need more open datasets like these to help researchers build speech systems that can serve many individuals around the world. 🗺️
MLCommons
@MLCommons
MLCommons releases Multilingual Spoken Words Corpus, permissively licensed keyword spotting dataset in 50 languages. The rich audio #speech dataset helps advance development of apps such as voice interfaces for a broad global audience. mlcommons.org/datasets/multilingual-spoken-words/ #ai #ml #neurips2021
-
Andrew Ng
@AndrewYNg
Thanks to everyone that participated in the Data-centric AI NeurIPS workshop! I was surprised and delighted at the sheer amount of innovation in the field. I also share some of my reflections from the workshop in The Batch. read.deeplearning.ai/the-batch/issue-122/
-
Andrew Ng
@AndrewYNg
We're holding the first NeurIPS Data-centric AI workshop next week on Dec 14th! We had over 160 paper submissions and I’m excited to hear about all this incredible work in Data-centric AI. Check out the program here: datacentricai.org
Andrew Ng
@AndrewYNg
Thank you to all the workshop presenters, plus a shoutout to my co-organizers @laroyo @CodyColeman2000 @GregoryDiamos @profvjreddi @joavanschoren Jean-Carole Wu @realsharonzhou
-
Andrew Ng
@AndrewYNg
Congratulations @jack for the wonderful institution that you've built Twitter to be.
-
Andrew Ng
@AndrewYNg
Thank you @FortuneMagazine @brianbokeefe for having me at the Brainstorm AI conference to talk about Data-centric AI and Landing AI's new fundraise. Great event! fortune.com/2021/11/08/andrew-ng-data-centric-artificial-…
-
Andrew Ng
@AndrewYNg
CNBC wrote a nice piece on Landing AI’s $57M Series A. Thank you to @McRockCapital @ScottyMacVC for leading the round, as well as investors Insight Partners, Taiwania, Canada Pension Plan, Intel, Samsung, DRIVE Catalyst, Walsin Lihwa, and AI Fund! www.cnbc.com/2021/11/08/google-brain-founder-andrew-ng-ra…
-
Andrew Ng
@AndrewYNg
I’m thrilled Landing AI has closed a $57M Series A, for our Data-centric MLOps platform for computer vision. Industrial AI needs a different recipe than Internet companies. It's time to make cutting-edge AI fast and easy for anyone to use!
-
Andrew Ng
@AndrewYNg
For this week’s The Batch, I wrote up some Halloween+AI puns. Hope you like them; and, if you have another one, please reply to this tweet to share yours too! read.deeplearning.ai/the-batch/issue-115/
-
Andrew Ng
@AndrewYNg
I'm a big fan of Hugging Face. Thank you for working with the deeplearning.ai team on this!
Hugging Face
@huggingface
We're thrilled to partner with deeplearning.ai to create some great new content for their NLP Specialization on Coursera!
With this update, you can access exciting new material and lectures that cover the state of the art in NLP 🧑🏫
www.coursera.org/specializations/natural-language-processing -
Andrew Ng
@AndrewYNg
Thank you to everyone who participated in the Data-centric AI Competition. I’m excited to announce the winners! 🏆 Best Performance Overall goes to Divakar Roy, @innotescus & Synaptic-AnN. Most Innovative goes to Mohammad Motamedi, @JohnsonHKuan, and @GoDataDriven. (1/2)
Andrew Ng
@AndrewYNg
Honorable Mentions also to Pierre-Louis Bescond and KAIST-AIPRLab.
Congratulations to all the teams! I’m looking forward to learning about their amazing stories--stay tuned! @DeepLearningAI_ @LandingAI https-deeplearning-ai.github.io/data-centric-comp/ (2/2) -
Andrew Ng
@AndrewYNg
I wrote in The Batch about some myths on Data-centric AI. What did I miss---are there any other misconceptions you’ve come across and should be addressed? If so please reply to this to let me know! read.deeplearning.ai/the-batch/issue-111/
-
Andrew Ng
@AndrewYNg
How would you define Data-centric AI development? It'd be useful for all of us collectively to write a sharable definition. Here's my attempt: Data-centric AI is the practice of systematically engineering the data used to build AI systems. What do you think/any suggestions?
-
Andrew Ng
@AndrewYNg
It is exciting to see @QuantumScapeCo’s anode-free approach to building electric vehicle batteries. I am thrilled @LandingAI is getting to support your team @ironmantimholme with LandingLens and help with the manufacturing of a revolutionary transportation tech!
Tim Holme
@ironmantimholme
We’re always looking for ways to enhance and accelerate our battery development, so I’m excited to be working with @AndrewYNg and @LandingAI to apply deep learning visual inspection to improve materials quality at QS.
-
Andrew Ng
@AndrewYNg
I would love your thoughts on this: Data-centric AI is still an emerging field, but what do you think are the key pillars of data-centric AI? E.g., if you were reading a textbook on this nascent field, what are some major topics you’d like the book to include?
-
Andrew Ng
@AndrewYNg
I wrote a short piece for Time Magazine on Jensen Huang's inclusion in Time 100, on his visionary leadership of the GPU revolution leading to incredible impact. Congratulations Jensen, and thank you for your work!
TIME
@TIME
“With still emerging AI technologies creating an insatiable hunger for more computation, Huang’s team is well-positioned to keep driving technological advances for decades to come,” writes @AndrewYNg #time100 ti.me/3CdX9Kf
-
Andrew Ng
@AndrewYNg
Happy International Programmers' Day! This is celebrated on the 256th day of the year, which is today. ❤️ all of you writing or learning to write software for other people to use!
-
Andrew Ng
@AndrewYNg
Hmm, one day later and two research groups have already submitted 4-page articles.🤔 We're off to a fast start!🔥
Andrew Ng
@AndrewYNg
Announcing the NeurIPS Data-Centric AI workshop! This will be the premier event for academic research on Data-Centric AI. Deadline for research article submissions is Sep 30th. Check it out! datacentricai.org
-
Andrew Ng
@AndrewYNg
Announcing the NeurIPS Data-Centric AI workshop! This will be the premier event for academic research on Data-Centric AI. Deadline for research article submissions is Sep 30th. Check it out! datacentricai.org
-
Andrew Ng
@AndrewYNg
I wrote about my 102 year old grandfather's passing in this week's The Batch. To everyone that reached out to express love and support, thank you. ❤️ read.deeplearning.ai/the-batch/issue-106/
-
Andrew Ng
@AndrewYNg
Congratulations to @kiankatan and the whole @workera_ team on this milestone. @DeepLearningAI_ and @AI_Fund are proud to support Workera!
kian
@kiankatan
Happy to announce @workera_ 's Series A led by @NEA!
We’ve raised $16M to continue helping organizations and individuals identify and close their skills gap in a measurable way with our skills assessment and precision upskilling technologies.
techcrunch.com/2021/08/26/workera-ai-a-precision-upskilli… -
Andrew Ng
@AndrewYNg
I had fun chatting with @megynkelly. Hoping this will spark additional discussions around a balanced view of AI. My part starts at 1:05:20. podcasts.apple.com/us/podcast/the-benefits-and-dangers-of… @MegynKellyShow
-
Andrew Ng
@AndrewYNg
We’re organizing a NeurIPS Workshop on Data-centric AI, to be held Dec 14th. Stay tuned… official announcement coming soon, together with the call for papers (invitation to submit your research work)!
-
Andrew Ng
@AndrewYNg
With the fall of Afghanistan's government, women's rights took a big step backward today. This is a sobering reminder that the rise of human rights isn't inevitable, and we have to keep pushing to bolster them.
-
Andrew Ng
@AndrewYNg
This writeup on the Data-Centric AI competition has some greats tips on data-centric approaches to getting good performance. Thanks Pierre-Louis Bescond for sharing these insights! medium.com/geekculture/a-deep-dive-into-andrew-ng-data-ce…
-
Andrew Ng
@AndrewYNg
Interested in learning how to improve machine learning performance using a data-centric AI approach?
Join me and a great expert panel @ajratner
@joavanschoren @salwanurm @CarlosAlzate @DillonLaird for a live event on Aug 11 at 10 am PT! RSVP: www.eventbrite.com/e/data-centric-ai-real-world-approache… -
Andrew Ng reposted
Ligand
@Ligand_LGND
Ligand is leveraging the @LandingAI LandingLens™ visual inspection software platform to further empower the #omniab® and xPloration™ antibody discovery platforms #ai #computervision #antibody $LGND www.prnewswire.com/news-releases/landing-ais-landinglens-…
-
Andrew Ng
@AndrewYNg
Both business and technical leaders interested in AI should get involved in the Data-Centric AI movement now. This will lower the barrier to adoption for many valuable applications. Check out this Harvard Business Review article to learn more. @HarvardBiz hbr.org/2021/07/ai-doesnt-have-to-be-too-complicated-or-e…
-
Andrew Ng
@AndrewYNg
Mental health is one of the most important problems of our time, and it is exciting to see Woebot Health leading the way. Congratulations to the @HiWoebot team on this milestone!
Woebot Health
@WoebotHealth
We're excited to announce our $90M #capital raise, co-led by @JAZZ_VP and @Temasek, with participation from @BlackRock Private Equity Partners and @OwlVC, allowing us to continue advancing the next frontier of technology-enabled #mentalhealth. woebothealth.com/woebot-health-closes-90-million-series-b… #ai
-
Andrew Ng
@AndrewYNg
Tell me you work in AI without telling me you work in AI.
-
Andrew Ng
@AndrewYNg
I’ve been following the Data-Centric AI competition leaderboard with excitement. Right now Wei Jing is in the lead, followed closely by AryanTyagi. Bi2i and @svpino are tied for third place. Anyone want to take them on? https-deeplearning-ai.github.io/data-centric-comp/
-
Andrew Ng
@AndrewYNg
Nice writeup by @Analyticsindiam on MLOps and the panel discussion with @chipro @robert_crowe @rajatmonga @lmoroney @ryanCkeenan. We still need better tools and principles for MLOps to make building and deploying AI models repeatable and systematic. analyticsindiamag.com/does-mlops-live-upto-the-hype/
Andrew Ng
@AndrewYNg
You can also watch a recording of the panel discussion here: www.youtube.com/watch?v=Ta14KpeZJok And to learn more about the cutting-edge field of MLOps, you can take this specialization here: www.coursera.org/specializations/machine-learning-enginee…
-
Andrew Ng
@AndrewYNg
Next week on June 30, I’ll be with my Machine Learning Engineering for Production (MLOps) Specialization co-instructors @robert_crowe and @lmoroney, as well as @chipro and @rajatmonga, in a live event to talk about MLOps. Hope to see you there! mlep.eventbrite.com
-
Andrew Ng
@AndrewYNg
Good article by @GilPress describing Data-Centric AI, and why this approach is critical to helping many industries create valuable AI systems. www.forbes.com/sites/gilpress/2021/06/16/andrew-ng-launch…
-
Andrew Ng
@AndrewYNg
Happy Father’s Day to all my fellow dads, and to everyone else celebrating today with a father! ❤️
And lots of love to my amazing dad--I wouldn't be the person I am today without his love and support. More about him: www.ricemedia.co/culture-people-after-son-cofounded-cours… -
Andrew Ng
@AndrewYNg
Announcing the Data-Centric AI competition! I’m excited to invite you to participate in this new competition format, and see how you can improve an AI system only by refining the data it depends on! https-deeplearning-ai.github.io/data-centric-comp/
-
Andrew Ng
@AndrewYNg
I'm very glad that the U.S. is buying 500 million doses of the Pfizer vaccine to share with the world. 🇺🇸 I hope other countries with sufficient resources will also step up, and lets get the whole world vaccinated together.
-
Andrew Ng reposted
Patrick David
@pdquant
Finding good content on ML Ops is hard
this course with @AndrewYNg @TensorFlow @lmoroney @DeepLearningAI_ @robert_crowe covers everything from model baselines through to deployment!
#mlops #deeplearning
blog.tensorflow.org/2021/06/mlep-courses.html -
Andrew Ng
@AndrewYNg
Congrats @anbarth, @Shelbee_SE, Sireesha Muppala and @cfregly on launching the Practical Data Science Specialization in partnership between @DeepLearningAI_ and @awscloud. Scaling from idea to production is a critical skill for all ML practitioners. www.coursera.org/specializations/practical-data-science
-
Andrew Ng
@AndrewYNg
Love seeing the data-centric AI development movement growing! Starting in June, FourthBrain (online AI bootcamp and @AI_Fund portfolio company) will be teaching data-centric approaches to MLOps! www.fourthbrain.ai/mlops-and-systems
-
Andrew Ng
@AndrewYNg
Congratulations @andrewmccallum and the whole Open Review team! Y'all are doing great work supporting the research publication process.
Open Review
@openreviewnet
OpenReview just smoothly weathered the #neurips2021 deadline today, hardly breaking a sweat. Over 41k sessions, 35k active users (29k in final 2 hours), more PDFs submitted than we’re allowed to say, sub-second response time on most requests, CPU utilization never surpassing 15%.
-
Andrew Ng
@AndrewYNg
Would love your feedback on this: AI Systems = Code (model/algorithm) + Data. Most academic benchmarks/competitions hold the Data fixed, and let teams work on the Code. Thinking of organizing something where we hold the Code fixed, and ask teams to work on the Data. (1/2)
Andrew Ng
@AndrewYNg
Hoping this will more closely reflect ML application practice, and also spur innovative research on data-centric AI development. What do you think? (2/2)
-
Andrew Ng
@AndrewYNg
Thrilled deeplearning.ai’s Machine Learning Engineering for Production (MLOps) specialization is now available on Coursera! Learn to build & deploy AI systems that run in production, and develop your production-ready skills: www.coursera.org/specializations/machine-learning-enginee… @lmoroney @robert_crowe
-
Andrew Ng reposted
Mike Nash
@MikeNashX
Brilliant points made by @AndrewYNg - Need to shift focus from being model-centric to being data-centric. So true, data is the lifeblood. Right data + good code = best results.
More: www.youtube.com/watch?v=06-AZXmwHjo
www.deeplearning.ai/the-batch/issue-84/
#artificialintelligence #data @DeepLearningAI_ -
Andrew Ng
@AndrewYNg
Whoops, sharing this late, but I'm proud Landing AI is in the CB Insights list of the 100 most promising AI startups! landing.ai/news/landing-ai-named-to-the-2021-cb-insights-… @CBinsights
-
Andrew Ng
@AndrewYNg
Just updated the Deep Learning Specialization with the latest advances. This 2021 release goes up to Transformer Networks, and Prog Exercises are now in TF2. If you’re wondering whether to learn DL, this is a good time to jump in! Learn here: www.coursera.org/specializations/deep-learning
-
Andrew Ng
@AndrewYNg
Let us always remember George Floyd, and keep fighting to make our society fair and just. #blm
-
Andrew Ng
@AndrewYNg
Join @drfeifei, @curtlanglotz, and me for a fireside chat next week on April 29 to hear us talk about healthcare’s AI future: healthcaresaifuture.eventbrite.com
-
Andrew Ng
@AndrewYNg
I stopped by @coursera's campus for a dry run of the Coursera Conference, and the team surprised me with birthday cake! Thanks, ❤️ you all!!!
-
Andrew Ng
@AndrewYNg
This Sunday is my birthday! The best gift🎁 to me would be if you can watch this video and let me know what you think. www.youtube.com/watch?v=06-AZXmwHjo
Lets you and I work to shift AI from Model-Centric toward Data-Centric AI development, which will help many teams. -
Andrew Ng
@AndrewYNg
Nice writeup by @Analyticsindiam on why I think we should move from model-centric AI development (where the emphasis is on improving models) toward #datacentricai (where we systematically improve the data, using MLOps tools). analyticsindiamag.com/big-data-to-good-data-andrew-ng-urg…
-
Andrew Ng reposted
-
Andrew Ng
@AndrewYNg
When I think about what we can do to build a brighter future for our children, the most important thing is to create a foundation for education. Because education is knowledge. And knowledge is human progress.
-
Andrew Ng
@AndrewYNg
To everyone speaking privately or publicly to #stopasianhate, I want to say: Thank you.
-
Andrew Ng
@AndrewYNg
I’ve been thinking about how to accelerate how all of us build and deploy ML, and have some ideas I want to share. I hope you’ll join me on this interactive livestream next Wednesday to chat over some ideas! mlops0324.eventbrite.com
-
Andrew Ng reposted
Berci Meskó, MD, PhD
@Berci
1/3
Check out this simple visual on the deployment of A.I. in healthcare: This is exactly why I’m subscribed to ‘The Batch’, the newsletter of deeplearning.ai!
This visual truly helps us understand the transformation from human to A.I. in procedures. -
Andrew Ng
@AndrewYNg
Seeing these exuberant, cheering faces is a great reminder that Perseverance's technology success is also a human success. Way to go NASA!
NASA
@NASA
Touchdown confirmed. The #countdowntomars is complete, but the mission is just beginning.
-
Andrew Ng
@AndrewYNg
126 million miles away, Perseverance decelerated from 12,000 miles/hour to 0 miles/hour to land safely. Great job @NASA team!
-
Andrew Ng
@AndrewYNg
Bearing is using AI to streamline maritime shipping operations. The shipping industry is a good example of a large, older industry, where AI is well-suited to improve manual processes that have been around for decades. bearing.ai/press-releases/bearing-launch-release
-
Andrew Ng
@AndrewYNg
We just reached the milestone of 10.0% of the U.S. receiving at least one vaccine shot! 🎉 We still have to stay vigilant, but here's to hoping for continued progress against Covid for all nations and to defeating the pandemic. 😷
-
Andrew Ng
@AndrewYNg
In 2021, what do you think is the single most important problem that the AI community should be working on?
-
Andrew Ng
@AndrewYNg
I'm with @fchollet on this. There're some best-practices on creating and organizing data that experienced applied ML people use, but we still need to flesh out and widely disseminate these ideas. This will be key to getting more ML systems deployed.
François Chollet
@fchollet
ML researchers work with fixed benchmark datasets, and spend all of their time searching over the knobs they do control: architecture & optimization. In applied ML, you're likely to spend most of your time on data collection and annotation -- where your investment will pay off.
-
Andrew Ng
@AndrewYNg
Covid's viral spread showed the speed and power of exponential growth. When vaccines get the local reproductive rate <1, I hope we'll see the speed and power of exponential decay. This is something to look forward to... but until then, let's keep masking up!
-
Andrew Ng
@AndrewYNg
I just told my daughter that if we can have a woman VP in the White House, then she, too can do anything. Since she's 2 years old, she responded that she would like some chocolate.
-
Andrew Ng
@AndrewYNg
Here's to wishing President @JoeBiden and VP @KamalaHarris the best! 🇺🇸 No matter where you're reading this, I hope their success will also be your success. ❤️
-
Andrew Ng reposted
Pranav Rajpurkar
@pranavrajpurkar
Does higher performance on ImageNet translate to higher performance on medical imaging tasks?
Surprisingly, the answer is no!
We investigate their relationship.
Paper: arxiv.org/abs/2101.06871
@_alexke, William Ellsworth, Oishi Banerjee, @AndrewYNg @StanfordAILab
1/8 -
Andrew Ng
@AndrewYNg
Today is Martin Luther King Jr. day. After the violence in the U.S. Capitol on Jan 6th, I am struck by how far MLK advanced civil rights through nonviolent activism. Let us build on his legacy and keep finding peaceful ways to advance civil rights. kinginstitute.stanford.edu/encyclopedia/nonviolence
-
Andrew Ng reposted
Sharon Zhou
@realSharonZhou
Excited to share our #iclr2021 paper w/ CS & math depts @Stanford 🎊
Evaluating the Disentanglement of Deep Generative Models through Manifold Topology!
w/ @ericzelikman Fred Lu @AndrewYNg Gunnar Carlsson @StefanoErmon. Acknowledging @torbjornlundh Samuel Bengmark.
Thread 🧵 -
Andrew Ng
@AndrewYNg
Democracy has taken a lot of damage recently. Last Wednesday was a nadir for the U.S., and the path back will be long and hard. But I believe the country has reached a turning point. Here're my thoughts on the role AI can play to support democracy. blog.deeplearning.ai/blog/the-batch-propagandists-lie-abo…
-
Andrew Ng
@AndrewYNg
U.S. democracy is going through a dark day today with the Capitol building breached. But we will get through this, democracy will prevail, and we will learn from this and strengthen our nation against future assaults. 🇺🇸
-
Andrew Ng
@AndrewYNg
Congrats @OpenAI on really cool image generation result from a text+image prompt! Here's my favorite set of generated examples:
-
Andrew Ng
@AndrewYNg
Happy New Year, friends! 🎉 2020 was a tough year for many, so lets make sure to build, support each other, and lift everyone up in 2021.
-
Andrew Ng
@AndrewYNg
Today's The Batch includes pieces from AI leaders @robotsmarts, @drfeifei, Matthew Mattina, @harryshum, and @ilyasut on their hopes for AI in 2021. Take a look! blog.deeplearning.ai/blog/the-batch-new-year-wishes-from-…
-
Andrew Ng
@AndrewYNg
This week's The Batch gives an overview of some major AI themes and stories we saw in 2020. Check it out! blog.deeplearning.ai/blog/the-batch-biggest-ai-stories-of…
-
Andrew Ng reposted
Santiago
@svpino
Probably my favorite newsletter with exquisite content every week.
From @DeepLearningAI_, featuring @AndrewYNg.
"Essential news for deep learners". -
Andrew Ng
@AndrewYNg
What rules regarding publishing papers would be fair, when it relates to work done by researchers working for companies? I ask this question in this week's The Batch, and would love to hear your thoughts. blog.deeplearning.ai/blog/the-batch-new-coronavirus-treat…
-
Andrew Ng reposted
Dr. Jill Biden
@DrBiden
Together, we will build a world where the accomplishments of our daughters will be celebrated, rather than diminished.
-
Andrew Ng
@AndrewYNg
Dismayed at @WSJopinion article's insulting, baseless and senseless challenge to Dr. Jill Biden's credentials. Dr. Biden earned her doctorate studying the important problem of student retention in community colleges. We need more scholars like her. www.wsj.com/articles/is-there-a-doctor-in-the-white-house…
-
Andrew Ng
@AndrewYNg
I'm backstage, watching a livestream of my own talk right now. Even though many of us are working from home, it's nice for so many worldwide to be able to gather together to talk about AI! #neurips2020 #ml4h
-
Andrew Ng
@AndrewYNg
Climate change remains one of humanity's biggest challenges, and AI can play a role! @jeremy_irvin16 @hao_ssr & @realSharonZhou just presented 3 papers at the #neurips CCAI workshop:
stanfordmlgroup.github.io/projects/ognet
stanfordmlgroup.github.io/projects/solar
stanfordmlgroup.github.io/projects/forestnet
Andrew Ng
@AndrewYNg
Thanks also to collaborators Kemen Austin, @DJGagneDos, @jack_kelly, Ram Rajagopal, Rob Jackson, Zutao Yang, Ritesh Gautam, Mark Omara, @DescartesLabs, @KemenAustin, & students @ericzelikman, Cooper Raterink, @neelr11, @sonjajohnsonyu, Chris Cross, Sasankh Munukutla, Shawn Zhang
-
Andrew Ng
@AndrewYNg
Indeed! Thank you @mattlungrenMD for giving my 2 year old daughter (who does not actually read) a 1000+ page text with a nice note encouraging her to pursue radiology as a career. Totally appropriate gift for an illiterate toddler. 😀
Judy Gichoya
@judywawira
@mattlungrenMD gifted @AndrewYNg 's 2 year old daughter with this book .... because his most treasured physical possession is a paper from his father... Well she needs to be an Interventional radiologist so she doesn't get replaced by AI :) great #neurips #ml4h keynote
-
Andrew Ng
@AndrewYNg
I talk about some factors that influence AI startup success in today’s The Batch: www.deeplearning.ai/the-batch
Andrew Ng
@AndrewYNg
If you're an aspiring AI entrepreneur, please let us know how we can help you in our short survey: share.hsforms.com/1XNvN2RIXTs2JBj_63_zdpg3hul4 @DeepLearningAI_ @AI_Fund
-
Andrew Ng
@AndrewYNg
It's important for every government's civil service to understand AI. I'm encouraged by Canada's progress, and am grateful to the CSPS for having me!
Nassim
@nassimmj
Look who is talking @DigiAcademyCAN
Annual Digital Open Government Forum!
@AndrewYNg -
Andrew Ng
@AndrewYNg
When HLP (human level performance) on an ML task is low, rather than just setting out to build a system to beat HLP, sometimes we should instead review of how the labels are defined, and this will lead to better performance. More details in today's The Batch letter.
Andrew Ng
@AndrewYNg
You can also find this week's and previous editions of The Batch here: deeplearning.ai/thebatch
-
Andrew Ng
@AndrewYNg
I talk about AI+Manufacturing in today's edition of The Batch, and why it is time to rethink how we use Human Level Performance as a metric that we try to beat.
-
Andrew Ng
@AndrewYNg
Thank you @KamalaHarris for setting this wonderful example to my daughter Nova. As a father, I loved this thought.
Kamala Harris
@KamalaHarris
While I may be the first woman in this office, I will not be the last—because every little girl watching tonight sees that this is a country of possibilities.
-
Andrew Ng
@AndrewYNg
I'm relieved at the election outcome and grateful to so many that made it possible. Thank you to every election worker helping run a fair process. To the media that defended the truth. And to each of you that voted or spoke up for participating in this wonderful democracy. 🇺🇸
-
Andrew Ng
@AndrewYNg
This seems appropriate to retweet for Halloween. 🦸🏻
Zachary Lipton
@zacharylipton
***In the first thrilling installment, the Superheroes of Deep Learning find themselves up against an ordinary problem. Will their extraordinary abilities carry the day?***
New on Approximately Correct! @AndrewYNg @drfeifei approximatelycorrect.com/2020/10/26/superheroes-of-deep-l… -
Andrew Ng
@AndrewYNg
I wrote about LandingLens and its unique features in The Batch. AI has the problem of needing customization, which is why we need verticalized platforms like LandingLens to empower manufacturing & other domain experts to build and deploy AI models. landing.ai/platform/
-
Andrew Ng
@AndrewYNg
I'm excited to announce @landingAI's LandingLens, a visual inspection platform to help manufacturers build and deploy visual inspection systems. We need to lower the bar to AI! landing.ai/landing-ai-unveils-ai-visual-inspection-platfo…
-
Andrew Ng
@AndrewYNg
Fun article about my amazing father, who has completed 146 online courses! (I also talk about this in today's edition of @DeepLearningAI_'s The Batch.) www.msn.com/en-sg/news/singapore/after-his-son-co-founded…
-
Andrew Ng
@AndrewYNg
No we're not, Mr. President. Living in California, I will keep working to benefit all Americans and all people everywhere. I hope everyone in the Federal Government will do so too.
Donald J. Trump
@realDonaldTrump
California is going to hell. Vote Trump!
-
Andrew Ng
@AndrewYNg
What are your favorite AI technology breakthroughs or demos from the past year? Would love to hear particularly if there's something you think not many others have heard of yet.
-
Andrew Ng
@AndrewYNg
The GANs Specialization by @DeepLearningAI_ just launched on @Coursera! You’ll master foundational GAN concepts, build your own GAN to transfer image styles, implement state-of-the-art StyleGAN techniques, and more. @realSharonZhou Check it out: www.coursera.org/specializations/generative-adversarial-n…
-
Andrew Ng
@AndrewYNg
Come join @goodfellow_ian, @AnimaAnandkumar, Alexei Efros, @realSharonZhou and me at GANs for Good, an online panel, next Wednesday Sep 30th! This is to celebrate the launch of @DeepLearningAI_’s new GANs Specialization. You can sign up here: www.eventbrite.com/e/gans-for-good-tickets-121256079197
-
Andrew Ng
@AndrewYNg
@goodfellow_ian, @AnimaAnandkumar, Alexei Efros, @realSharonZhou and I will be speaking at GANs for Good, an online panel, on September 30th. This is to celebrate the launch of @DeepLearningAI_’s new GANs Specialization. Come join us! www.eventbrite.com/e/gans-for-good-tickets-121256079197
-
Andrew Ng
@AndrewYNg
Thanks to everyone that replied to my question on what the AI community should work on. I think if we come together as a community we can make better progress. I wrote about this in The Batch today (reposted here). Would love to hear your thoughts!
-
Andrew Ng
@AndrewYNg
More companies should tell employees it's fine if your kids show up on Zoom when you're working from home. This isn't a sign of someone being unprofessional. This is a sign that you're a human being.
-
Andrew Ng
@AndrewYNg
What do you think is the single most important problem that the AI community should be working on?
-
Andrew Ng
@AndrewYNg
Ken Church, Marti Hearst, @lukaszkaiser, @YMourri & I will speak at Break into NLP, a virtual live event on July 29 at 10am PDT, to mark the launch of Course 3 of the @deeplearningai_ NLP Specialization. Come join us! deeplearningainlp.eventbrite.com
-
Andrew Ng reposted
Coursera
@coursera
Coursera’s #globalskillsindex 2020 is here! See how your country and industry rank in essential skills across business, technology, and data science. www.coursera.org/skills-reports/global
#skills4tomorrow -
Andrew Ng
@AndrewYNg
Wonderful! Kudos to Harvard and MIT for fighting this in court, as well as to everyone that spoke up against the new ICE rules.
The Harvard Crimson
@thecrimson
Breaking: The government has agreed to rescind DHS and ICE rules barring international students attending online universities from staying in the U.S., per a hearing this afternoon in Harvard and MIT's lawsuit against the agencies.
-
Andrew Ng
@AndrewYNg
I'm grateful for Dr. Fauci's tireless work to disseminate good science about the pandemic.
When we see a scientist unfairly attacked, we need to speak up. It is by doing so that we defend not just the scientist, but also the scientific process.
#istandwithfauci -
Andrew Ng
@AndrewYNg
I wrote in today's edition of The Batch about the new F-1 visa policy, and want to share that message here as well. #studentban
-
Andrew Ng
@AndrewYNg
New @ICEgov policy regarding F-1 visa international students is horrible & will hurt the US, students, and universities. Pushes universities to offer in-person classes even if unsafe or no pedagogical benefit, or students to leave US amidst pandemic and risk inability to return.
Andrew Ng
@AndrewYNg
Here's the text of the policy. This puts the US, students and universities in a lose-lose-lose situation. www.ice.gov/doclib/sevis/pdf/bcm2007-01.pdf
-
Andrew Ng reposted
Matt Sarmiento
@msarmi9
@AndrewYNg’s the batch is consistently my favourite weekly newsletter and one I always can’t wait to get my hands on.
It’s absolutely fantastic and one I highly recommend to any and all 🙃
www.deeplearning.ai/thebatch/ -
Andrew Ng
@AndrewYNg
The suspension of the H1B visa program is bad for the US, bad for innovation, and will shatter dreams and disrupt lives. As a former H1B visa holder, my heart goes out to all the families affected.
-
Andrew Ng reposted
Pranav Rajpurkar
@pranavrajpurkar
New publication!
We apply machine learning to predict response to antidepressant treatments in patients with major depressive disorder @JAMANetworkOpen
jamanetwork.com/journals/jamanetworkopen/fullarticle/2767367
w/ @JingboEYang, Nathan Dass, Vinjai Vale, @AndrewYNg, Leanne Williams and more.
Tweet thread...1/n -
Andrew Ng
@AndrewYNg
Happy Juneteenth! Raising awareness this past month on Black lives matter was important. But what’s next is even more important: that we keep supporting it even years from now when it’s not a trending twitter topic. Racial justice will take years; lets be in it for the long haul.
-
Andrew Ng
@AndrewYNg
Congrats @karpathy for being named a TR35 recipient! www.technologyreview.com/innovator/andrej-karpathy/ You can also listen to his interview with @deeplearningai_ here: www.deeplearning.ai/blog/hodl-andrej-karpathy/
-
Andrew Ng
@AndrewYNg
Our new Natural Language Processing Specialization’s first two courses are now on @Coursera! You’ll build your technical foundations and work on projects ranging from sentiment analysis to language modeling to autocomplete. Check it out:
www.coursera.org/specializations/natural-language-processing -
Andrew Ng
@AndrewYNg
Congratulations the #stanford2020 class that just graduated today! An online commencement wasn't what anyone had envisioned, but I am excited to see what you will accomplish and the contributions you'll make to this chaotic world. Proud of all of you! commencement.stanford.edu
-
Andrew Ng
@AndrewYNg
Justice is indivisible. Injustice anywhere is a threat to justice everywhere. -MLK
-
Andrew Ng reposted
Coursera
@coursera
University & college students can now learn in-demand career skills for free with full access to Coursera’s 3,800 courses, Guided Projects, & Professional Certificates. Find out more and help us tell the world! #skilluptogether
-
Andrew Ng
@AndrewYNg
Getting to widespread AI adoption outside tech companies means solving new challenges. For example, we still need to get better at bridging Proof of Concept (PoC) and production. Here are my thoughts: @FortuneMagazine fortune.com/2020/04/26/ai-data-big-data-manufacturing/
-
Andrew Ng
@AndrewYNg
I'm really enjoying the new #iclr2020 website, and am having fun browsing the workshop & poster talks. Congrats to the organizers and speakers on what's happening so far in this all-virtual conference! Great job! @srush_nlp @shakir_za @kchonyc @dawnsongtweets @syhw et al.
-
Andrew Ng
@AndrewYNg
My grandparents had many stories of helping neighbors during World War II, a defining time for them. When some day we are telling our own grandchildren about World Pandemic I, I hope all of us will have stories of our actions during this time that we can be proud of.
-
Andrew Ng
@AndrewYNg
I've listened to several of Shakira's songs over the years. So I was excited to see she just finished a UPenn course on @coursera! www.instagram.com/p/B_VCDIxqXfN/
-
Andrew Ng
@AndrewYNg
Social distancing is needed to keep us safe. So @LandingAI created this AI tool to help our customers in essential industries maintain #socialdistancing in the workplace. We hope many people can benefit from these ideas. landing.ai
-
Andrew Ng
@AndrewYNg
One of the fastest-growing AI applications is medicine. I'm excited to announce Courses 1&2 of @deeplearningai_‘s new AI For Medicine Specialization are now available on @Coursera! You’ll learn to diagnose from X-rays and build your own prognostic models: www.coursera.org/specializations/ai-for-medicine
-
Andrew Ng
@AndrewYNg
Many people are going through a hard time--anxious about the situation, financial worries, loneliness--because of Covid-19 and social distancing. I’m really glad that the @HiWoebot team has been working to help. woebot.io/blog/covid-19/
-
Andrew Ng
@AndrewYNg
Kudos to @jack for his $1B donation for Covid-19 efforts. He's one of the people I really admire--in addition to being a skilled executive, he's also a decent human being that cares about others. Today's world needs more leaders like Jack, and I'm glad he's leading Twitter.
-
Andrew Ng
@AndrewYNg
In anticipation of @deeplearningai_‘s new AI For Medicine Specialization, I’ll be speaking at our online event with @EricTopol, @lhpeng, and @pranavrajpurkar about this fast-growing application area. You can register here: ai4m.eventbrite.com
-
Andrew Ng reposted
GrepMed
@grepmeded
Great article. We need to raise our standards and learn from countries more successful at controlling this virus.
medium.com/@covid19workinggroup/how-hospitals-can-protect…
Andrew Ng
@AndrewYNg
Fighting COVID-19 requires protecting healthcare workers. The mask shortage is widely reported, but that’s not the full story--other practices are needed. We synthesized lessons from various countries that are useful for the US: medium.com/@covid19workinggroup/how-hospitals-can-protect…
-
Andrew Ng
@AndrewYNg
Fighting COVID-19 requires protecting healthcare workers. The mask shortage is widely reported, but that’s not the full story--other practices are needed. We synthesized lessons from various countries that are useful for the US: medium.com/@covid19workinggroup/how-hospitals-can-protect…
Andrew Ng
@AndrewYNg
With Ken Jung, @ronlivs, @RoseLikeTheFlwr, @EricTopol, Kelly Zhang, Ming Zhou, @realSharonZhou
-
Andrew Ng
@AndrewYNg
Congratulations to my colleague Pat Hanrahan and to Ed Catmull on winning the Turing Award for work in Computer Graphics, the effects of which we now enjoy in many movies. Fun fact: ~12 years ago my students were also getting advice from Pat's group on GPGPU for DL!
-
Andrew Ng
@AndrewYNg
What tips do you have for remote work? Tech has adopted remote ahead of most others. Lets do our part to help others implement remote where possible, to speed up #socialdistancing. Please reply and share your ideas for #remotework. #covidー19
-
Andrew Ng
@AndrewYNg
Happy Pi Day! 🥧We announced the Pie & AI series one year ago, and in 52 weeks have hosted 64 Pie & AI events in 32 countries, thanks to our event ambassadors. @deeplearningai_
Andrew Ng
@AndrewYNg
With COVID-19, I hope everyone will continue to organize meetings but online, and keep sharing knowledge. Lets support each other through this!
-
Andrew Ng reposted
Coursera
@coursera
We’re committed to helping educators minimize disruption during the COVID-19 outbreak. Starting today, with the support of our partners, we’ll provide every impacted university with free access to our course catalog through Coursera for Campus. Learn more: blog.coursera.org/helping-universities-and-colleges-go-fu…
-
Andrew Ng
@AndrewYNg
1/2 Google just announced their new TensorFlow Developer Certificate, which is a great way to showcase your TF skills. Check it out: tensorflow.org/certificate
Andrew Ng
@AndrewYNg
2/2 If you want to take the cert but don’t know TF yet, @deeplearningai_ @coursera’s TensorFlow in Practice specialization is the best way to prepare for the exam: www.coursera.org/professional-certificates/tensorflow-in-…
-
Andrew Ng
@AndrewYNg
A huge thank you to everyone in healthcare---doctors, nurses, hospital staff, epidemiologists, scientists, and others---working to fight the Covid-19 pandemic and protect all of us.
-
Andrew Ng
@AndrewYNg
Our latest on video pre-training for CT diagnosis: CT scans give 3D volumetric outputs & can be represented as video of different slices through the body. We pretrain w/500K YouTube videos, then transfer/finetune w/~400 CT images. Here's @pranavrajpurkar's tweetstorm summary
Pranav Rajpurkar
@pranavrajpurkar
Tweet summary of our new work, AppendiXNet, on deep learning for diagnosis of appendicitis.
Key idea: video pretraining for CT.
www.nature.com/articles/s41598-020-61055-6
@bpatel_mba @allisonjypark @jeremy_irvin16
@c_a_chute @curtlanglotz @mattlungrenMD @AndrewYNg @StanfordAILab
Thread 1/10 -
Andrew Ng
@AndrewYNg
Chatting with @lexfridman about ML, rise of MOOCs, what’s hard in learning AI, ... was the most fun interview I’ve done recently. (Thanks Lex for such a great conversation!) You can watch it here: www.youtube.com/watch?v=0jspaMLxBig
-
Andrew Ng
@AndrewYNg
Consumer internet companies with 1B+ users have popularized Big Data. But industries like manufacturing, agriculture & healthcare need AI to work on Small Data. Good overview of emerging Small Data techniques, by @landingAI's @AlejoBt. www.industryweek.com/technology-and-iiot/digital-tools/ar…
-
Andrew Ng
@AndrewYNg
Happy multi-cultural palindrome day! Works whether your preferred date format is MM/DD/YYYY or DD/MM/YYYY or YYYY/MM/DD.
-
Andrew Ng
@AndrewYNg
Great chatting with @lexfridman. We talked about everything from the early mistakes I made building MOOCs to the challenges of deploying AI systems. Stay tuned for the podcast episode!
-
Andrew Ng
@AndrewYNg
Great having @EricTopol visit just now. His book Deep Medicine gives a compelling vision for AI in healthcare. We had fun chatting about ML applied to everything from EKG to X-rays to EHR, and how exciting it is to see widespread progress on many fronts!
-
Andrew Ng
@AndrewYNg
RIP Professor Clayton Christensen. The world has lost a brilliant mind and a wonderful person, whose teachings on the Innovator's Dilemma have influenced countless people, and are as relevant as ever for navigating AI's disruptive impact. We will miss you.
-
Andrew Ng reposted
Sayak Paul
@RisingSayak
The latest @deeplearningai_ newsletter is just amazing. The part that I liked the most is the #imagenet dataset is being modified and efforts are being made to purge the bias outta it.
Read the jam-packed newsletter here: blog.deeplearning.ai/blog/the-batch-facebook-takes-on-dee…. -
Andrew Ng
@AndrewYNg
Many aspiring AI engineers ask me how to take the next step and join an AI team. This report from @Workera_, a @deeplearningai_ affiliate, walks you through how AI teams work and which skills you need for different AI career tracks. Download it here: www.workera.ai/candidates/report
-
Andrew Ng
@AndrewYNg
In The Batch's new year edition, @ylecun @katecrawford @kaifulee @animaanandkumar @etzioni @chelseabfinn @RichardSocher @dawnsongtweets, Dave Patterson & Zhi-Hua Zhou share their hopes for AI in 2020. Check it out (and subscribe if you haven't yet!): www.deeplearning.ai/the-batch
-
Andrew Ng
@AndrewYNg
Happy New Decade! In the last decade AI went from niche to mainstream. I wonder what our community will accomplish this decade?
-
Andrew Ng
@AndrewYNg
Do you have an example of an underrated or underreported AI result from 2019--something that deserves to be more widely known? Please reply and share your thoughts!
-
Andrew Ng reposted
Climate Change AI
@ClimateChangeAI
Submissions for our ICLR workshop "Tackling Climate Change with Machine Learning" are now open!
Papers and Proposals due: Feb 4
Apply to mentor or be mentored on submissions: Jan 14
More details: www.climatechange.ai/ICLR2020_workshop -
Andrew Ng
@AndrewYNg
Reflecting on #neurips2019. Great meeting as usual, but the AI community is maturing, which felt bittersweet. Bitter because:
Post not available here, view on X
Andrew Ng
@AndrewYNg
It was fun speaking on the climate change panel with Yoshua Bengio, @JeffDean, Carla Gomes, and Lester Mackey. Thanks @david_rolnick, @priyald17, @LynnHKaack, and others for organizing the great workshop.
Andrew Ng
@AndrewYNg
Overall, a great conference. Congratulations and thanks to @hannawallach, @hugo_larochelle, Alina Beygelzimer, Florence d'Alché-Buc, Emily Fox, @jennwvaughan, @DaniCMBelg, @aliceoh and others for putting this together.
-
Andrew Ng
@AndrewYNg
Courses 1 & 2 of @deeplearningai_‘s newest Specialization is now available on @Coursera! Training a model is only one step in building a working AI system. These courses teach you how to navigate some key deployment scenarios. Enroll here: www.coursera.org/specializations/tensorflow-data-and-depl…
-
Andrew Ng
@AndrewYNg
I’m excited about the opening of @Stanford’s new hospital! In addition to improving patient care, hopefully will also be a new platform for many AI and healthcare innovations. Congrats to all that made this possible! @DEntwistleSHC med.stanford.edu/news/all-news/2019/11/stanford-health-ca…
-
Andrew Ng
@AndrewYNg
Since announcing Pie & AI in March, we’ve held events in 16 cities around the world. Let's keep the momentum going! If you want to host an event with us and support your local community of deep learners, you can apply at deeplearning.ai/events.
-
Andrew Ng
@AndrewYNg
It’s Halloween, but we don’t have to succumb to our fears, especially around the future of AI. If you want to know how scared you should or shouldn’t be, you can read our take on 6 AI fears in the Halloween special edition of The Batch! info.deeplearning.ai/the-batch-daemon-spawn-agi-takeover-…
-
Andrew Ng
@AndrewYNg
I recently reread Fred Brooks’ classic essay “No Silver Bullet,” as I’ve been thinking about how to reduce the essential complexity of building ML systems. Shared some thoughts about it in The Batch this week, which you can read here: info.deeplearning.ai/the-batch-google-achieves-quantum-su…
-
Andrew Ng reposted
DeepLearning.AI
@DeepLearningAI
Watch @AndrewYNg at @TechCrunch Sessions: Enterprise at 4PM PT! Catch the livestream here: www.youtube.com/watch?v=Nkyu8DLtdjA&hss_channel=tw-992153…
-
Andrew Ng reposted
TechCrunch
@TechCrunch
Andrew Ng’s AI companies expand to Medellin, Colombia tcrn.ch/2Zl2MGM by @fredericl
-
Andrew Ng
@AndrewYNg
Announcing our first international office in Medellin, Colombia! Landing AI, deeplearning.ai and AI Fund have teams there. AI needs to expand beyond a small handful of hubs like Silicon Valley and Beijing, and I’m bullish about Latin America. medium.com/@andrewng/ai-in-latin-america-announcing-our-f…
-
Andrew Ng
@AndrewYNg
Introducing The Batch, a new weekly newsletter from @deeplearningai_! AI is growing so fast that it’s hard to know what news is useful or valuable. Subscribe to The Batch for a curated report on the latest AI research and industry-shaping events: www.deeplearning.ai/the-batch
-
Andrew Ng
@AndrewYNg
Congratulations to Martial Hebert on becoming CMU's new Dean of CS! I have fond memories of working with him, @SebastianThrun, and others long ago on a DARPA project using helicopters+ground robots to build 3D maps. Looking forward to more great things from @SCSatCMU!
-
Andrew Ng
@AndrewYNg
The final course of the deeplearning.ai TensorFlow Specialization is now available on Coursera! You’ll learn how to build time series models, and at the end you’ll train a deep neural network on real-world data to predict sunspot activity: www.coursera.org/professional-certificates/tensorflow-in-…
-
Andrew Ng
@AndrewYNg
Computer Scientist Alan Turing will appear on UK currency. This is wonderful! Perhaps this acknowledgement of his accomplishments will encourage more people to enter CS and AI? www.bbc.com/news/business-48962557
-
Andrew Ng
@AndrewYNg
Many companies outside the software industry are still in the early phases of adopting AI. Landing AI’s @DongyanWang8 talks about selecting the right initial projects, and avoiding an initial fumble that saps momentum. @venturebeat venturebeat.com/2019/07/12/to-be-successful-with-ai-you-h…
-
Andrew Ng
@AndrewYNg
My friend @erikbryn just stopped by the office. I always find his work on AI's impact on the economy enlightening; this work is much needed in the "new electricity" era. Also had a great conversation about Conditional Basic Income and the future of education!
-
Andrew Ng
@AndrewYNg
I’m excited to welcome @KellySeelig as our new VP of Marketing & Communications. Having built brands for some of Silicon Valley’s most notable companies, Kelly’s already igniting the team with fresh ideas. Check out our new website! landing.ai
-
Andrew Ng
@AndrewYNg
I’m glad DeepNude is dead. As a person and as a father, I thought this was one of the most disgusting applications of AI. To the AI Community: You have superpowers, and what you build matters. Please use your powers on worthy projects that move the world forward.
-
Andrew Ng
@AndrewYNg
Course 3 of the deeplearning.ai TensorFlow Specialization is now available on Coursera! You’ll learn how to process text using tokenization and train LSTMs to create original poetry. Enroll in the Specialization for $49/month or audit for free: www.coursera.org/professional-certificates/tensorflow-in-…
-
Andrew Ng
@AndrewYNg
My fun activity this morning: @KellerRinaudo just gave me a video tour of their Ghana @zipline facility. In the pic you can see the drones (sans wings) in the background, which they use to ship medical supplies to remote hospitals. Exciting use of AI+Drones!
-
Andrew Ng
@AndrewYNg
What a great day! Just wrapped up @Stanford's 128th commencement. Congratulations to all of our graduates. We're proud of you and look forward to hearing about your future accomplishments!
-
Andrew Ng
@AndrewYNg
Had a great time at the Amazon #remars conference last week. Even got to take a ride in the @blueorigin capsule. Thanks to @astro_pat and Dean Kassmann for showing me around!
-
Andrew Ng
@AndrewYNg
Our new result published in @JAMANetworkOpen showing ML+ radiologist outperforms a human radiologist alone at detecting cerebral aneurysms! news.stanford.edu/2019/06/07/ai-tool-helps-radiologists-d…
@allisonjypark, @c_a_chute, @pranavrajpurkar, @RobynLBall, @mattlungrenMD, @KristenYeom @Stanford -
Andrew Ng
@AndrewYNg
Looking forward to the coming AI transformation of radiology, and to building it with global radiologist community. Thank you @curtlanglotz, @mattlungrenMD & Udo Hoffman for having @pranavrajpurkar and me present at the @RSNA course on AI!
-
Andrew Ng
@AndrewYNg
Today we honor the men and women that gave everything in service, fighting for everyone's freedom. 🇺🇸 #memorialday
This is an emotional day for many of the fallen's family and loved ones. Spare a thought for them too, and to them I say also: Thank you. -
Andrew Ng
@AndrewYNg
Great conversation with Christine Payne @mcleavey about her work on MuseNet music composition. A year ago, she took the Deep Learning Specialization @deeplearningai_ and now she’s working on cutting-edge NNs at @OpenAI!
This media couldn't be displayed. See it on the original post. -
Andrew Ng
@AndrewYNg
Hanging out with the Landing AI team at #makerfaire today. Rainy but hundreds of cool robots! Also had a few ppl ask me about adding AI to their projects. 😀
-
Andrew Ng
@AndrewYNg
For the Deep Learning Specialization on Coursera, I interviewed some of the greats of DL. I hope they inspire you to start studying DL yourself: www.coursera.org/specializations/deep-learning
This article has key takeaways from the Heroes of Deep Learning interviews: www.analyticsvidhya.com/blog/2018/09/heroes-deep-learning… -
Andrew Ng
@AndrewYNg
Woebot just won the Google Play Award in the Wellbeing App category! I’m super excited about what Woebot, as a mental health chatbot, can do to help you in a moment of need. @alisonmdarcy @HiWoebot 9to5google.com/2019/05/06/2019-google-play-award-winners/
-
Andrew Ng
@AndrewYNg
1/On #worldpressfreedomday, I’d like to thank all the reporters working hard to tell important stories, whether in tech or politics or something else. For democracy to function and make good decisions, we need the truth to be known.
Andrew Ng
@AndrewYNg
2/If there’s a newspaper or magazine whose reporters you respect, I also urge you to subscribe and pay for it. These dollars are crucial for funding the press and enabling reporters to do their important work.
-
Andrew Ng
@AndrewYNg
RIP to my friend, colleague, and AI visionary Nils Nilsson. Your work on the A* algorithm has improved countless lives (this is how we find the shortest path from A to B). I will always remember your work, but even more importantly your kindness. ai.stanford.edu/~nilsson/
-
Andrew Ng
@AndrewYNg
I’m happy to announce my lab is working on AI+Climate Change. Climate Change is one of humanity’s most pressing problems & the Tech community must help. I’m still exploring more projects. If you want to help, let me know here: deeplearning.ai/climate #earthday
Andrew Ng
@AndrewYNg
Further, my @Stanford group is running an AI+Climate Change bootcamp with climate faculty. We’ve held an AI+CC symposium at Stanford (more details later) and you can also see more info on the bootcamp here.
stanfordmlgroup.github.io/programs/aicc-bootcamp/ -
Andrew Ng
@AndrewYNg
Course 2 of the @deeplearningai_ @TensorFlow Specialization is now available on @Coursera! You’ll learn how to use TF to handle real-world data, avoid overfitting w/augmentation and dropout, and more. You can take the course for $49 or audit it for free: www.coursera.org/learn/convolutional-neural-networks-tens…
-
Andrew Ng
@AndrewYNg
My @Stanford lab just released the MRNet dataset of 1000+ annotated knee MRIs. Join our competition to develop and test your knee MRI interpreting deep learning model: stanfordmlgroup.github.io/competitions/mrnet @nbien2 @pranavrajpurkar @RobynLBall @jeremy_irvin16 @curtlanglotz @mattlungrenMD
-
Andrew Ng
@AndrewYNg
My former student Ian Goodfellow moving to Apple is the big ML news of today. Best of luck @goodfellow_ian, and congrats @tim_cook for getting him! www.i-programmer.info/news/105-artificial-intelligence/12…
-
Andrew Ng
@AndrewYNg
Wrapping up another @Coursera Partners Conference! I remain super grateful to our incredible partners for creating amazing courses for everyone. Also fun to see AI applications to edu are accelerating. Lively debates about AI autograders, analytics & more! #courseracon2019
-
Andrew Ng
@AndrewYNg
Great chatting with @demishassabis and @koraykv. Thanks for sharing the latest at @DeepMindAI. You all are doing inspiring work!
-
Andrew Ng
@AndrewYNg
Saying bye to my daughter Nova before I head to London for a few days.
I have an irrational worry that by the time I get home on Thursday she'll be all grown up and have forgotten who I am. That's not how it works, right? -
Andrew Ng
@AndrewYNg
Hanging out with @lmoroney to plan deeplearning.ai TensorFlow specialization materials. More videos are coming soon!
Hey Laurence: cool coffee mug. ☕️😀 -
Andrew Ng
@AndrewYNg
Imperial College @imperialcollege is launching an online Masters degree in Machine Learning with @coursera. This is exciting! Given the huge global demand for machine learning engineers, the world needs a lot more degree programs in AI and ML.
www.imperial.ac.uk/news/190523/imperial-launches-worlds-f… -
Andrew Ng
@AndrewYNg
About time! Geoff Hinton, @ylecun and Yoshua Bengio were just selected for the ACM Turing Award for their Neural Network work going all the way back to the 1980s. Congrats to all three! This is a great step for all of AI!
Andrew Ng
@AndrewYNg
If you want to hear their stories behind these decades of work, also check out my interviews with them on @coursera as part of the Deep Learning specialization:
www.coursera.org/lecture/neural-networks-deep-learning/ge…
www.coursera.org/lecture/convolutional-neural-networks/ya…
www.coursera.org/lecture/deep-neural-network/yoshua-bengi… -
Andrew Ng
@AndrewYNg
I’m thrilled to welcome Dongyan Wang as Landing AI’s new VP AI Transformations! He led strategic AI/IT initiatives at Midea Group, NetApp and Cisco. At Landing AI, he will lead customer engagement and strategic partnerships to help other enterprises adopt AI! @DongyanWang8
-
Andrew Ng
@AndrewYNg
Woebot’s CEO @alisonmdarcy did a cool interview on the @ShouldThisExist podcast. Woebot has more conversations in a week than a therapist will have in their entire career! itunes.apple.com/us/podcast/should-this-exist/id145021552…
-
Andrew Ng
@AndrewYNg
While having pie for Pi Day last Thursday, inspiration struck. We enjoy AI, we enjoy pie, and we want to share both with more people. 🥧
Introducing Pie & AI: a @deeplearningai_ meetup series. Sign up so you’ll know when we’re coming to your town! deeplearning.ai/events -
Andrew Ng
@AndrewYNg
One of the (many) reasons AI research is progressing faster than ever is that there're now more people doing AI research than at any other point in history.
This is a wonderful trend, and lets keep welcoming more people into AI! -
Andrew Ng reposted
Amazon News
@amazonnews
Save the date: Registration for #remars, Amazon’s new #ai event, opens March 28. Featured speakers include @JeffBezos, @RobertDowneyJR, @AndrewYNg and more. Head to our blog for agenda and registration information. www.aboutamazon.com/innovation/how-ai-will-shape-the-futu…
-
Andrew Ng
@AndrewYNg
Happy Pi Day everyone!
Trying to read a research paper, but surprisingly hard to focus when wondering what flavor of pie we're getting. 🥧 -
Andrew Ng
@AndrewYNg
New result from our @Stanford lab! Instead of giving everyone the same treatment, you can use ML to predict who would benefit from high blood pressure treatment and who wouldn't. @tonyduan_ @pranavrajpurkar @DillonLaird, and Sanjay Basu. Details here: www.ahajournals.org/doi/full/10.1161/CIRCOUTCOMES.118.005010
-
Andrew Ng
@AndrewYNg
It was great catching up with @CarnegieMellon's Prof @Tommmitchell at the Landing AI office. I've know him since my undergrad days at CMU, and we had fun reminiscing on what @deanpomerleau's 1994 Neural Networks course was like!
Andrew Ng
@AndrewYNg
Thanks to Tom's work on the CMU face recognition dataset, this gif of me now exists on the internet forever. 😎
-
Andrew Ng
@AndrewYNg
Announcing deeplearning.ai's newest specialization, TensorFlow: From Basics to Mastery! In partnership with @TensorFlow team. To be great at implementing AI, you need to know how to best use ML frameworks like TF. @deeplearningai_ Take Course 1: www.coursera.org/learn/introduction-tensorflow/
-
Andrew Ng
@AndrewYNg
I'm very excited to welcome @tedgreenwald to the @deeplearningai_ team! Ted is a former @WSJ editor, and will be leading a new editorial function to share with you the most important stories in AI. Stay tuned!
-
Andrew Ng reposted
Kousuke Ariga
@kouariga
A list of @AndrewYNg's AI intros for non-technical people. I recommend going through all of them if you are interested in starting a business using AI.
10mins: hbr.org/2019/02/how-to-choose-your-first-ai-project
30mins: landing.ai/ai-transformation-playbook/
2hrs: www.youtube.com/watch?v=21EiKfQYZXc
A month: www.coursera.org/learn/ai-for-everyone -
Andrew Ng
@AndrewYNg
AI for Everyone was designed as a 4 week course, but 252 learners have already finished the whole thing! Congrats to all 252 of you. 🎉 I don't recommend "cramming" everything into one day, but it is possible to do the whole thing quickly if you want.
-
Andrew Ng
@AndrewYNg
AI For Everyone is now available on @Coursera! This course will help non-engineers and engineers work together to leverage AI capabilities and build an AI strategy. If you want your company to embrace AI, this is the course to get your CEO to take! coursera.org/learn/ai-for-everyone
-
Andrew Ng
@AndrewYNg
1/We’re one day from the “AI for everyone” launch! My secret plan is that by having more people know how to work with AI engineers and build AI projects, that this will open up even more AI jobs and let there be more AI in the world.
Andrew Ng
@AndrewYNg
2/Okay now that I’ve tweeted this out I guess it’s not secret anymore.
-
Andrew Ng
@AndrewYNg
AI For Everyone will launch on February 28th! This non-technical course will teach you the language of AI, how to drive AI adoption in your company, and AI’s potential impact on society. @deeplearningai_ @Coursera You can pre-enroll here: coursera.org/learn/ai-for-everyone
-
Andrew Ng
@AndrewYNg
Companies are hiring data privacy experts. This is a good step! But privacy (similar to security) should not be the sole domain of privacy engineers. In addition to the experts, it's also every engineer's responsibility to do what they can. fortune.com/2019/02/20/data-privacy-jobs-in-demand/
-
Andrew Ng reposted
NOVA | PBS
@novapbs
Great name 😉
-
Andrew Ng
@AndrewYNg
Announcing Nova Ng, our first daughter. Plus, some thoughts on the AI-powered world she will grow up in: medium.com/@andrewng/nova-ng-38c9fb2883bb
-
Andrew Ng reposted
Christopher Bishop
@chrisbishop
Andrew Ng: How to Choose Your First #ai Project hbr.org/2019/02/how-to-choose-your-first-ai-project via @AndrewYNg @HarvardBiz terrific guidance - must read for anyone investigating how to get started generating business value from #artificialintelligence
-
Andrew Ng
@AndrewYNg
Did you know the Universal Declaration of Human Rights says "higher education shall be equally accessible to all on the basis of merit"? The crushing US college debt is a sobering reminder we still have much work ahead to make education accessible. www.bloomberg.com/news/articles/2019-02-16/u-s-student-de…
-
Andrew Ng
@AndrewYNg
Back in 2010, thanks to this figure generated by @adampaulcoates, I went to Larry Page to pitch using More Data + Bigger Models for NNs, which became Google Brain (w/@JeffDean). Looking at OpenAI's new NLP result, we've not yet reached the limits of NNs at scale.
Andrew Ng
@AndrewYNg
For context, here's the paper (published in 2011). I wouldn't use this algorithm today though! ai.stanford.edu/~acoates/papers/coatesleeng_aistats_2011.pdf
-
Andrew Ng
@AndrewYNg
If your company wants to ramp up in AI, how should you choose your first few projects? I just wrote a Harvard Business Review @HarvardBiz article with practical suggestions. This fleshes out the first step from the AI Transformation Playbook. hbr.org/2019/02/how-to-choose-your-first-ai-project
-
Andrew Ng
@AndrewYNg
Gopi Prashanth @gopiprshnth is one of the unsung heroes of AI. Among other things, he built the AI tech for Amazon Go--one of the most amazing AI products of our generation. I'm thrilled to welcome him as Landing AI's new VP of Engineering! venturebeat.com/2019/02/06/landing-ai-hires-amazon-direct…
-
Andrew Ng
@AndrewYNg
The @WSJ gives a summary of the AI Transformation Playbook published by Landing AI. The playbook synthesizes my learnings from building/leading AI teams and speaking with many CEOs, and I hope helps many companies adopt AI. @SCastellWSJ www.wsj.com/articles/reporters-notebook-andrew-ngs-tips-f…
Andrew Ng
@AndrewYNg
You can read the AI Transformation Playbook here landing.ai/ai-transformation-playbook/
-
Andrew Ng
@AndrewYNg
Welcome @Walmart to the self-driving world, and congratulations on your visually distinctive car design. 😉
I guess orange really is the new black! @driveai_ 🏎️🏎️
news.walmart.com/2019/01/08/custom-autonomous-cargo-vans-… -
Andrew Ng
@AndrewYNg
Self-driving cars will be one of the most important AI applications in the world. Want to study this topic? Check out the Self-Driving Car specialization on @coursera, offered by @UofT! venturebeat.com/2019/01/30/coursera-and-the-university-of…
-
Andrew Ng
@AndrewYNg
Pictures from our new office! I love our new place. Though still not as much as I love getting to work with all these great people on AI! ❤️
-
Andrew Ng
@AndrewYNg
Announcing CheXpert! Large dataset of chest X-rays co-released with MIT's MIMIC-CXR dataset. Join our competition to test your chest X-ray interpretation model: stanfordmlgroup.github.io/competitions/chexpert
@jeremy_irvin16 @pranavrajpurkar @mlko53 @curtlanglotz @mattlungrenMD @Stanford #aaai19 -
Andrew Ng
@AndrewYNg
The whole team is packing to move into our new office. I’d signed a 36 month lease for our current place, and we outgrew it in 10 months. In hindsight, I didn’t project our headcount growth well at all. (Maybe I should have used deep learning. 💁🏻♂️)
-
Andrew Ng
@AndrewYNg
A little peek at our latest project. Baby Ng is on the way - Arriving February 2019! @robot_MD
-
Andrew Ng reposted
Woebot Health
@WoebotHealth
TOP of the charts!!! Our app is trending #1! Check us out on
Android: play.google.com/store/apps/details?id=com.woebot
iOS: itunes.apple.com/us/app/woebot-depression-anxiety/id13053…
#android #ios #playstore #appstore #mentalhealth #mentalhealthawareness #ai #healthcare #healthtech #wellness -
Andrew Ng
@AndrewYNg
To every US government employee: A huge thank you. I am saddened that so many of you are unfairly asked to work without pay or be furloughed. But I am very grateful for your service to the nation. 🇺🇸
-
Andrew Ng
@AndrewYNg
It was a pleasure speaking with @mfordfuture for his new book, Architects of Intelligence. This is a great book for anyone who wants to better understand how AI will impact our economy and society. book.mfordfuture.com
-
Andrew Ng
@AndrewYNg
Do you have any suggestions for what to name our 8 new conference rooms? If so please reply to this tweet! Landing AI, @deeplearningai_ and AI Fund have grown so fast we've moving into a bigger space this month.
-
Andrew Ng
@AndrewYNg
Landing AI is thrilled to enter into partnership with LG Electronics. Looking forward to working closely with Dr. I.P. Park and his team! #ces2019 www.lgnewsroom.com/2019/01/lg-and-landing-ai-join-forces-…
-
Andrew Ng
@AndrewYNg
Backstage at #ces2019 with the LGE team. Looking forward to going onstage in a few minutes!
-
Andrew Ng
@AndrewYNg
Our new paper in Nature Medicine! Cardiologist-level arrhythmia detection from ECG, using deep learning. www.nature.com/articles/s41591-018-0268-3 I'm very optimistic about such technology helping patients.
@awnihannun, @pranavrajpurkar, @leftbundle, @GeoffTison @iRhythmTech
Andrew Ng
@AndrewYNg
More info also on our project homepage: stanfordmlgroup.github.io/projects/ecg2/
-
Andrew Ng
@AndrewYNg
1/The rise of Software Engineering required inventing processes like version control, code review, agile, to help teams work effectively. The rise of AI & Machine Learning Engineering is now requiring new processes, like how we split train/dev/test, model zoos, etc.
Andrew Ng
@AndrewYNg
2/I'm also seeing many AI teams use new processes that haven't been formalized or named yet, ranging from how we write product requirement docs to how we version data and ML pipelines. This is an exciting time for developing these ideas!
Andrew Ng
@AndrewYNg
3/Have you seen an idea for organizing ML projects that you'd like to share with others? If so please reply to this tweet!
-
Andrew Ng
@AndrewYNg
We cannot abdicate responsibility when two children, ages 7 and 8, die in US custody. The US once said: "Give me your tired, your poor, your huddled masses yearning to breathe free." I have faith in the US to be better than this.
Donald J. Trump
@realDonaldTrump
Any deaths of children or others at the Border are strictly the fault of the Democrats and their pathetic immigration policies that allow people to make the long trek thinking they can enter our country illegally. They can’t. If we had a Wall, they wouldn’t even try! The two.....
-
Andrew Ng
@AndrewYNg
Introducing the AI Transformation Playbook! How can your company become good at AI? This 5-step Playbook provides the key steps for transforming your company using AI, from executing pilot projects to building a team, and more.
Download your free copy: landing.ai/ai-transformation-playbook/ -
Andrew Ng
@AndrewYNg
The AI Index 2018 report is out! Lots of great data. My key takeaways: (i) AI's rapid growth--in jobs, publications, performance--continues. (ii) We still need to do better in diversity/inclusion. @indexingai cdn.aiindex.org/2018/AI Index 2018 Annual Report.pdf
-
Andrew Ng
@AndrewYNg
Announcing AI For Everyone, the newest @deeplearningai_ course! Whether you're a CEO, product manager, marketer, recruiter, designer, or financier, this non-technical course will help you understand how to apply AI in your company, and navigate AI's rise. medium.com/@andrewng/announcing-ai-for-everyone-a-new-cou…
-
Andrew Ng
@AndrewYNg
Apparently 20 years ago I was already contributing to face recognition research. 😃 #tbt
DeepLearning.AI
@DeepLearningAI
Look who showed up in the 1997 CMU Face Dataset 😎#tbt @AndrewYNg
Props to @ricardoaraujo for unearthing this! -
Andrew Ng
@AndrewYNg
The new UN Climate report shows our planet is nearing crisis. While individuals tweaking their carbon footprint is a good step, we now need more than that. The best hope for our planet is electing decisive global leaders who will make the right choices. Please vote.
-
Andrew Ng
@AndrewYNg
AI+ethics is important, but has been partly hijacked by the AGI (artificial general intelligence) hype. Let's cut out the AGI nonsense and spend more time on the urgent problems: Job loss/stagnant wages, undermining democracy, discrimination/bias, wealth inequality.
-
Andrew Ng
@AndrewYNg
Podcast on building AI businesses and AI's societal impact, w/ @GreylockVC Greymatter. Thanks @saranormous for doing this with me! soundcloud.com/greylock-partners/greylock-andrewsarah-pod…
-
Andrew Ng
@AndrewYNg
Congratulations to my friend @lilaibrahim for her new role as COO of Deepmind! She had played a pivotal role as Coursera's Chief Operations Officer, and I'm happy to see her now playing a bigger role in AI as well. deepmind.com/blog/welcoming-lila-ibrahim-our-first-chief-…
-
Andrew Ng
@AndrewYNg
Stephen Hawking once said, "I'm not afraid of death, but I'm in no hurry to die. I have so much I want to do first." RIP Professor Hawking, and may we all strive to live as fully as he did.
-
Andrew Ng
@AndrewYNg
Do you want to learn about AI+Healthcare? Our @Stanford team is releasing a new Medium publication to share exciting work across the field! Check it out: medium.com/stanford-ai-for-healthcare
-
Andrew Ng
@AndrewYNg
If someone builds a Deep Learning-powered startup, should we call them entrepreneurial or entrepreneural?
-
Andrew Ng
@AndrewYNg
The AI Fund will be building companies from the ground up. We'll also work to bring more people into AI to do electrifying work! techcrunch.com/2018/01/30/andrew-ng-officially-launches-h…
-
Andrew Ng
@AndrewYNg
1/Need to time technology well: 2007 was good time to launch iPhone; but not 1993 (Apple Newton) since battery/screen/chip tech not there.
Post not available here, view on X
Andrew Ng
@AndrewYNg
8/Adding to this thread 5 months later. Here's what I think is now great timing for AI: Manufacturing! www.technologyreview.com/s/609770/andrew-ng-says-factorie…
-
Andrew Ng
@AndrewYNg
The Wall Street Journal just joined the debate on whether radiologists should worry about their jobs. @greg_ip www.wsj.com/articles/how-robots-may-make-radiologists-job…
-
Andrew Ng
@AndrewYNg
Here's our full paper on improving palliative care with deep learning. Neural net with 18 hidden layers inputs EHR, estimates chance of mortality in next 3-12 months. Also generates report to explain predictions to clinicians. arxiv.org/abs/1711.06402
-
Andrew Ng
@AndrewYNg
Very cool Kaggle survey on state of data science. www.kaggle.com/surveys/2017
-
Andrew Ng
@AndrewYNg
Stanford's first day of class--record-breaking 1040 people already enrolled for on-campus Machine Learning (CS229). Wow! @danboneh
-
Andrew Ng
@AndrewYNg
Kudos Mark Zuckerberg for speaking out against AI fear mongering. fortune.com/2017/07/25/mark-zuckerberg-elon-musk-artifici…
-
Andrew Ng
@AndrewYNg
Technical details on our Deep Learning+ECG (detecting irregular heartbeats/arrhythmia) work: stanfordmlgroup.github.io/projects/ecg/
-
Andrew Ng
@AndrewYNg
Launching my new project! Hope will help many of you: deeplearning.ai More announcements soon. #deeplearningai
-
Andrew Ng
@AndrewYNg
Mobile flower/etc. recognition has done well in China since 2014. Nice to see this coming to the US through Google Lens at I/O.
-
Andrew Ng
@AndrewYNg
If you're trying to understand AI's near-term impact, don't think "sentience." Instead think "automation on steroids."
-
Andrew Ng
@AndrewYNg
I will be resigning from Baidu and opening up a new chapter of my work in AI. Details here: medium.com/@andrewng/opening-a-new-chapter-of-my-work-in-…
-
Andrew Ng
@AndrewYNg
I often advise people to take on projects you're only 70% qualified for, but then learn like crazy to bridge that 30%.
-
Andrew Ng
@AndrewYNg
You know you're a nerd when coolest thing at Superbowl is not the game, ads, or Lady Gaga, but the fleet of autonomous drones.
-
Andrew Ng
@AndrewYNg
Speak up on twitter if you haven't yet about #muslimban. #deleteuber influenced Travis Kalanick. Every voice, including yours, matters.
-
Andrew Ng
@AndrewYNg
CMU just made history: AI beats top humans at Texas Hold'em poker. A stunning accomplishment, comparable to Deep Blue & AlphaGo!
-
Andrew Ng
@AndrewYNg
Loved @claychristensen's "Competing against luck." Tips to ensure your org is doing job your customer wants. www.amazon.com/Competing-Against-Luck-Innovation-Customer…
-
Andrew Ng
@AndrewYNg
MOOC completion predicts tech interview performance more than what school attended. Amazing! h/t Leith Abdulla
blog.interviewing.io/lessons-from-3000-technical-interviews/ -
Andrew Ng
@AndrewYNg
When was the last time you (or any other human) encountered the trolley dilemma in real life? Expect self-driving car to see it as often.
-
Andrew Ng
@AndrewYNg
The half-life of knowledge is decreasing. That's why you need to keep learning your whole life, not only through college.
-
Andrew Ng
@AndrewYNg
Many S&P 500 CEOs wish the had an internet strategy sooner. In 5 years we'll hear same about AI strategy. fortune.com/2016/10/05/ai-artificial-intelligence-deep-le…
-
Andrew Ng
@AndrewYNg
Check out our new voice-centric keyboard: play.google.com/store/apps/details?id=com.baidu.research.… TalkType- type 3x faster using high-accuracy speech!
-
Andrew Ng
@AndrewYNg
35,000 have signed up for Machine Learning Yearning book. Join us by Jun 24th! mlyearning.org
-
Andrew Ng
@AndrewYNg
Honored to be named to the CNN 10: Thinkers list, together with Elon Musk and others! bit.ly/HfDb5I
-
Andrew Ng
@AndrewYNg
Just named to 2013 #time100 (World's most influential people)! ti.me/YXSHKP






















































































































































































































