Today we're thrilled to announce the release of our first frontier visual reasoning model
Ian Goodfellow
Ian J. Goodfellow (born 1987) is an American computer scientist, engineer, and executive, most noted for his work on artificial neural networks and deep learning. He was a research scientist at Google DeepMind, was previously employed as a research scientist at Google Brain, and director of machine learning at Apple, as well as one of the first employees at OpenAI. He has made several important contributions to the field of deep learning, including the invention of the generative adversarial network (GAN). Goodfellow co-wrote, as the first author, the textbook Deep Learning (2016) and wrote the chapter on deep learning in the authoritative textbook of the field of artificial intelligence, Artificial Intelligence: A Modern Approach (used in more than 1,500 universities in 135 countries). More on Wikipedia
-
Ian Goodfellow reposted
Elorian AI
@ElorianAI
This media couldn't be displayed. See it on the original post. -
Ian Goodfellow
@goodfellow_ian
In Aisle's first year, they've shown that the system built around an LLM is highly important, and even small, cheap fast models can be boosted to do valuable security work
Stanislav Fort
@stanislavfort
My AI security research lab AISLE discovered 500+ zero-days with CVEs across 100+ codebases in a year => ~1% of 2026's open-source CVEs globally from a single team! On hardened codebases like curl, we've repeatedly matched or outperformed frontier AI labs in zero-day discovery
-
Ian Goodfellow reposted
Stanislav Fort
@stanislavfort
OpenAI's Codex Security and Anthropic's Mythos both reported 0 remaining security issues in curl
AISLE then discovered 6 new zero-days that curl's security team validated & has now assigned CVEs to
6-to-0 on one of the world’s most hardened codebases:
aisle.com/blog/aisle-discovered-six-curl-cves-after-opena… -
Ian Goodfellow
@goodfellow_ian
New BBC World Service documentary: "The Rise of Deepfakes: How AI Changed Porn, Politics, and Reality" youtu.be/JBBeNQ8L6FI
-
Ian Goodfellow
@goodfellow_ian
New office decoration: a GAN-dalf from @DhuviKarthikey1
-
Ian Goodfellow
@goodfellow_ian
I'm excited for the launch of @224ventureshq . I'm one of their LPs, helping to identify exceptional founders early
shaunbjohnson
@shaunbjohnson
-
Ian Goodfellow
@goodfellow_ian
While Mythos has popularized the idea of finding vulnerabilities with LLMs, Aisle was doing it earlier. From an engineering point of view, it's interesting to see that a small, open weight model with a structured search system is competitive at this task
Stanislav Fort
@stanislavfort
We matched Mythos on public zero-days with CVEs using widely available & open-source derived models & can run it air-gapped if needed. All this with a small team out of Europe
Berkeley study ranks us #1 globally in 3 of 8 categories
The full evidence: stanislavfort.substack.com/p/mythos-at-home-and-its-calle… -
Ian Goodfellow
@goodfellow_ian
Hi everyone, please donate to TechAide's fundraiser, organized by Hugo Larochelle. This is a good cause, I was one of the TechAide speakers two years ago.
Hugo Larochelle
@hugo_larochelle
Looking forward to our TechAide AI Conference today, during which we're raising funds to fight poverty and social exclusion in Montreal.
If you'd like to contribute, you can make a donation here:
centraidemtl.org/TechAide-AI-Conference-2026-DD
Any amount is greatly appreciate!! -
Ian Goodfellow
@goodfellow_ian
My former Brain colleague @AndrewDai just launched @ElorianAI
Andrew M. Dai
@AndrewDai
After almost 12 years in Brain/DeepMind, I’ve finally decided to take the leap. My cofounders: @yinfeiy, Seth and I have kicked-off @ElorianAI. The first multimodal reasoning lab founded and led by former LLM pretraining, data and multimodal leads. youtu.be/YlvfNpOMeOY (1/n)
-
Ian Goodfellow reposted
Public Health Action Network
@pubhealthaction
March 31st is the last day to submit proposals for projects with measurable results in reducing indoor airborne pathogens, like breath-based multi-pathogen detection systems, continuous HVAC compliance verification systems, and indoor air infrastructure deployment.
-
Ian Goodfellow
@goodfellow_ian
In Public Health Action Network we’re about to go through an internal expert review process to choose how to focus our efforts over the next few years. We’ll be deciding, how can we, as a small non-profit, most effectively reduce the spread of airborne disease. If you have creative, high leverage ideas for us to pursue, please submit a proposal for us to evaluate alongside our own ideas.
Public Health Action Network
@pubhealthaction
PHAN is now accepting requests for proposals for projects that will have measurable results in reducing transmission of indoor airborne pathogens. These could be projects like breath-based multi-pathogen detection systems or continuous HVAC compliance verification systems.
-
Ian Goodfellow
@goodfellow_ian
An article from Moonlake about why they’re building what they’re building. (I’m an advisor)
-
Ian Goodfellow
@goodfellow_ian
I'd like to thank @daniel_rossett for his help in my recovery from the POTS version of Long COVID. Daniel was key in bringing me back from highly disabled and suffering to being able to do what I want to again.
This X account is mostly focused on ML / AI. From that point of view, many of you know that in December 2024, I wasn't able to do the test of time award talk at NeurIPS, even by video call. Daniel started working
with me in March 2025. By April, I started to have days of no POTS symptoms, by June I was off all heart rate lowering medications, by September I was back to work. I'm back to full exercise, running, lifting weights, mountain biking, and have even done things I hadn't done before I got sick, like riding Whistler Mountain Bike Park.
I'm now getting the word out to help Daniel build a company that will bring this approach to more people. -
Ian Goodfellow
@goodfellow_ian
Amazing test of Gemini 3’s multimodal reasoning capabilities: try generating a threejs voxel art scene using only an image as input
Prompt: I have provided an image. Code a beautiful voxel art scene inspired by this image. Write threejs code as a single-page
Ian Goodfellow
@goodfellow_ian
Best outputs from Gemini 2.5 Pro, vs 3 Pro, this example nicely illustrates the fidelity jump with 3 Pro, through strong multimodal understanding and 3D reasoning
Ian Goodfellow
@goodfellow_ian
Try it for yourself at aistudio.google.com/apps/bundled/image_to_voxel
-
Ian Goodfellow
@goodfellow_ian
More of what my colleagues and I have been working on in AI for fusion is public now
Google DeepMind
@GoogleDeepMind
We’re announcing a research collaboration with @CFS_energy, one of the world’s leading nuclear fusion companies.
Together, we’re helping speed up the development of clean, safe, limitless fusion power with AI. ⚛️ -
Ian Goodfellow reposted
Yoshua Bengio
@Yoshua_Bengio
AI is evolving too quickly for an annual report to suffice. To help policymakers keep pace, we're introducing the first Key Update to the International AI Safety Report. 🧵⬇️
(1/10) -
Ian Goodfellow
@goodfellow_ian
Moonlake is launching publicly today, focused on generating simulations and games. Simulations are an important application of generative models that I’ve advocated for since 2014
Moonlake
@moonlake
We raised $28M seed from Threshold Ventures, AIX Ventures, and NVentures (Nvidia's venture capital arm) —alongside 10+ unicorn founders and top AI researchers— to build reasoning models that generate real-time simulations and games.
Models are bottlenecked by practical simulations that can act as Reinforcement Learning environments. Human self-expression is bounded by tools that let us create alternate realities.
At Moonlake, we are building a future where anyone can create interactive worlds, bring their child-like wonder to life, learn within them, and most importantly, share experiences with people we care about.
More in 🧵This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
Yoshua Bengio
@Yoshua_Bengio
It was an honour to speak at the @UN this week to address the UN Security Council on the impacts of AI on international peace & security, and to join the high-level multi-stakeholder informal meeting to launch the Global Dialogue on AI Governance.
We need global cooperation to steer AI toward a safe and beneficial future, and I was very pleased to see so many people come together these past few days with that shared goal. -
Ian Goodfellow
@goodfellow_ian
Check out @snowglobe_so , the new simulation engine for testing chatbots from @guardrails_ai
shreya rajpal
@ShreyaR
Introducing ❄️ @snowglobe_so, the simulation engine for AI chatbots.
Magically simulate the behavior of your users to test and improve your chatbots.
Find failures before your users do.This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
Demis Hassabis
@demishassabis
Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! deepmind.google/discover/blog/advanced-version-of-gemini-…
-
Ian Goodfellow
@goodfellow_ian
My account was compromised and now recovered. I have deleted a fraudulent post about a crypto token using the DeepMind name. Please do not participate if you saw that while it was up.
-
Ian Goodfellow reposted
Scott Kupor
@skupor
Great to see the power of hallucination-free voice AI by @InfinitusAI in driving up quality and lower costs for health care, starting with health risk assessments. www.fiercehealthcare.com/ai-and-machine-learning/infinitu…
-
Ian Goodfellow reposted
Public Health Action Network
@pubhealthaction
Hi Twitter friends! Want to make a better world for those with COVID and #longcovid? Or a world where everyone has clean indoor air? We want YOU! We need volunteers for our projects and communications work. Fill out our contact form on our website or DM us, and we'll be in touch!
-
Ian Goodfellow
@goodfellow_ian
This is personal for these founders, who are making this app to support important people in their own lives. They’ve also been reaching out for feedback from other chronic illness patients including me
Danielh Carranza
@carranzadanielh
Introducing CONY - Your Personal Health Companion
Cony sees your health holistically. Labs, wearables, symptoms, medications, nutrition, activity, medical records, environment–and how everything connects.
Cony, proactively monitors your health, answers your questions accurately, and recommends plans that adapt to your changes in real-time.
Cony acts as your personal health scientist, constantly researching, designing plans, and making discoveries about you.
Over time, it learns patterns about you and identifies clinically significant findings to generate reports for clinicians, facilitating proactive care.
Please comment to try the beta!This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
Ror Preston
@RorPreston
This chart shows that long COVID is now amongst the most burdensome diseases in the USA 😳
Thanks to @mildTin for her analysis and DALY estimate for long COVID 💙
Worth noting, this includes COVID-triggered ME/CFS -
Ian Goodfellow
@goodfellow_ian
Be sure to take our survey!
Public Health Action Network
@pubhealthaction
Hi Twitter friends! Have you taken our survey about our planned global resources directory? The directory will have information on COVID-conscious businesses, organizations, healthcare providers, and lots of other resources, and we'd love your input! publichealthactionnetwork.org/phan-safer-survey/
-
Ian Goodfellow
@goodfellow_ian
Benchmarks are both difficult and important. Great to see these thoughtful, open reproducible benchmarks for LLM guardrails.
shreya rajpal
@ShreyaR
📣Announcing Guardrails Index!!🎉
We’ve come a long way since we first launched @guardrails_ai in 2023, with guardrails now a pretty standard part of AI platforms everywhere
BUT how performant are guardrails really?!
I’m CONSTANTLY surprised by the lack of benchmarks for this -
Ian Goodfellow reposted
shreya rajpal
@ShreyaR
📣Announcing Guardrails Index!!🎉
We’ve come a long way since we first launched @guardrails_ai in 2023, with guardrails now a pretty standard part of AI platforms everywhere
BUT how performant are guardrails really?!
I’m CONSTANTLY surprised by the lack of benchmarks for this -
Ian Goodfellow reposted
Liesl McConchie
@LieslMcConchie
🚨Know Your Rights!🚨
Clean Air Allies (the non-profit I co-founded w/ @cleanairschool) is launching our 1st campaign.
Did u know California students have a constitutional right to safe & healthy school facilities?
Here's what you need to know & do⬇️
🧵
cleanairallies.org -
Ian Goodfellow
@goodfellow_ian
If you’re interested in resources for staying safe from COVID, take our survey on how Public Health Action Network can best support you: publichealthactionnetwork.org/phan-safer-survey/
Ian Goodfellow
@goodfellow_ian
LDN has helped me to stop having PEM. I recommend starting at a very low dose like .5mg, increasing slowly and only when you’re sleeping ok. I regret following instructions to work toward 4.5mg and getting stuck at a higher dose than I’d like (I feel sick if I try to reduce it).
-
Ian Goodfellow reposted
Mark Blair | Technosociologist
@mblair
There isn't enough research money behind it solving Long Covid/MECFS.
We are leaving people on the battlefield of the pandemic.
It should be receiving at least as much funding as HIV is today, considering that we already have treatments for that. -
Ian Goodfellow
@goodfellow_ian
100% this. My kids are missing out on the majority of the benefit of having a dad because of Long COVID (mostly POTS for me) and it’s right at the age where they need one most
Darren Parkinson 🌍💚🤘
@DarrenBar88
Not being able to do things with the kids is a biggy.
I miss this so much.
From family bike rides, to walks, to seeing them perform their music.
I feel like I'm not being the parent I should be. It's so difficult. -
Ian Goodfellow
@goodfellow_ian
Sign up for our newsletter!
Public Health Action Network
@pubhealthaction
Hi Twitter friends! We're putting together our first monthly email newsletter to go out tomorrow, and if you haven't subscribed yet, you should! Sign up on our website: publichealthactionnetwork.org/contact/
#covid #longcovid #stillcoviding #covidisntover #cleanair #cleanairforall -
Ian Goodfellow
@goodfellow_ian
The recording of the GAN test of time talk by @dwf is now publicly available: neurips.cc/virtual/2024/test-of-time/105032
-
Ian Goodfellow
@goodfellow_ian
My co-authors have completed one of the projects I was working on with them before I went on disability leave. Check out this great explainer thread from @davlindner
David Lindner
@davlindner
New Google DeepMind safety paper! LLM agents are coming – how do we stop them finding complex plans to hack the reward?
Our method, MONA, prevents many such hacks, *even if* humans are unable to detect them!
Inspired by myopic optimization but better performance – details in🧵 -
Ian Goodfellow
@goodfellow_ian
South Bay people, are you able to see your GP in less than 2 months? Mine at PAMF almost always cancels same day and reschedules me for ~6 weeks later
-
Ian Goodfellow reposted
Dr. Phillip Alvelda
@alvelda
Wow. In failing to address or abate the ongoing COVID pandemic, we are damaging our own cognition 3x worse than the Roman Empire inflicted upon itself with brain damage from lead pipes.
The big difference? They had no idea what heavy metal poisoning could do in 100 AD.
We’ve known of brain damage from COVID since 2021. Our failure to halt the pandemic doesn’t reflect well on us, does it?
Danielle Beckman, PhD
@DaniBeckman
"Widespread use of metal caused estimated 2- to 3-point drop in IQ for nearly 180 years of Pax Romana"
Meanwhile, #covid in 5 years:
"Participants with persistent symptoms had a 6-point loss in IQ, while those who had been admitted to an ICU experienced a 9-point loss in IQ." -
Ian Goodfellow
@goodfellow_ian
Schmidhuber was a reviewer of the GAN paper in 2014. The reviews are public. He is the reviewer that asked us to cite predictability minimization. He did not ask us to cite artificial creativity (his current complaint). He hadn’t thought of that even in 2016.
Ian Goodfellow
@goodfellow_ian
When Schmidhuber interrupted my talk on GANs in 2016, he did so at the time I was giving credit to relevant prior work: noise contrastive estimation. Schmidhuber isn’t about proper about academic credit assignment, he’s about self-aggrandizement.
-
Ian Goodfellow reposted
Ian Goodfellow
@goodfellow_ian
When Schmidhuber interrupted my talk on GANs in 2016, he did so at the time I was giving credit to relevant prior work: noise contrastive estimation. Schmidhuber isn’t about proper about academic credit assignment, he’s about self-aggrandizement.
-
Ian Goodfellow
@goodfellow_ian
When Schmidhuber interrupted my talk on GANs in 2016, he did so at the time I was giving credit to relevant prior work: noise contrastive estimation. Schmidhuber isn’t about proper about academic credit assignment, he’s about self-aggrandizement.
-
Ian Goodfellow reposted
hardmaru
@hardmaru
Brings back some memories from 2016🍿
hardmaru
@hardmaru
NIPS2016 Day 1: Poor @Goodfellow_Ian gets Schmidhuber'ed during educational GAN Tutorial session.
-
Ian Goodfellow
@goodfellow_ian
Today GANs got 1 of 2 test of time awards at @NeurIPSConf . Many thanks to my co-author @dwf for preparing and presenting the talk. It’s no easy task to summarize 85k papers in 12 minutes
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
If you’d like to help prevent the spread of COVID and other diseases, consider donating, volunteering, or applying to work at the nonprofit that I cofounded: www.publichealthactionnetwork.org
Ian Goodfellow
@goodfellow_ian
Congratulations to @dwf for a great talk. I thought he did a great job of giving credit to all the lab members that made GANs possible and describing how LISA was the right environment to create them as well as summarizing the impact they had after publication.
-
Ian Goodfellow
@goodfellow_ian
Another more specific way people can help: does anyone know if it’s safe to take betahistine if I have MCAS? And does taking antihistamines make betahistine ineffective?
Ian Goodfellow
@goodfellow_ian
Posting a call for help: does anyone know of a good way to simultaneously treat both POTS and Ménière’s disease? Please contact me if you’re either a clinician with experience doing this or a patient who has found a good solution. Context in thread
-
Ian Goodfellow
@goodfellow_ian
I’ll add what I’ve tried here, let’s try to get critical mass
-
Ian Goodfellow
@goodfellow_ian
My GAN co-author Sherjil Ozair has written about some memories of 2012-2014 in the context of GANs winning one of this year's test of time awards, worth a read for the nostalgia if you were around back then, or for learning what it was like if you weren't
Sherjil Ozair
@sherjilozair
Very happy to hear that GANs are getting the test of time award at NeurIPS 2024.
The NeurIPS test of time awards are given to papers which have stood the test of the time for a decade.
I took some time to reminisce how GANs came about and how AI has evolve in the last decade. -
Ian Goodfellow reposted
Friesein
@Friesein
The time has come for us to recognize aerosol hygiene as a necessity for modern life.
To do this, we need engineering controls that don't rely on human behavior.
Just as our ancestors decided it was time to filter the water, the same is now true of the air.
It's time to adapt. -
Ian Goodfellow
@goodfellow_ian
Posting a call for help: does anyone know of a good way to simultaneously treat both POTS and Ménière’s disease? Please contact me if you’re either a clinician with experience doing this or a patient who has found a good solution. Context in thread
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
The huge problem for me is that POTS and Meniere’s have basically opposite management requirements. Diuretics are very bad for POTS so I can’t take a high enough dose to keep my Meniere’s stable
Ian Goodfellow
@goodfellow_ian
With unstable Meniere’s, my ears pressurize and lose hearing if I lower my head, so I have to sleep propped up quite high, but this is horrible for POTS, since my body struggles to get blood into my head. Obviously this is bad for my sleep generally and causing a wider downward spiral.
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
My Meniere’s doctors are great but they don’t know a lot about POTS or POTS-compatible treatments. I’ve had less access to POTS doctors and they don’t know much about Meniere’s. I’m really struggling to find a way to manage both. Any help finding someone who knows what to do is greatly appreciated.
Ian Goodfellow
@goodfellow_ian
A few ideas for leads would be if there is anyone with experience treating Meniere’s with amiloride or fludrocortisone, or anyone with experience treating POTS patients who have to take diuretics for more common reasons than Meniere’s.
-
Ian Goodfellow
@goodfellow_ian
Looking for just a few hundred more signatures by tomorrow
#MEAction Network
@MEActNet
The last day to sign our letter to the NIH calling on the director to fund the ME/CFS Research Roadmap is this Friday, Nov. 15th!
US signers: win.newmode.net/fundmeroadmap
International signers: airtable.com/appls0UcwWjmI3TWw/pagMvv9RZYLxZiDFe/form
It will take all of us to show that we are united!
#pwme -
Ian Goodfellow reposted
#MEAction Network
@MEActNet
The last day to sign our letter to the NIH calling on the director to fund the ME/CFS Research Roadmap is this Friday, Nov. 15th!
US signers: win.newmode.net/fundmeroadmap
International signers: airtable.com/appls0UcwWjmI3TWw/pagMvv9RZYLxZiDFe/form
It will take all of us to show that we are united!
#pwme -
Ian Goodfellow
@goodfellow_ian
I’ve known Rohan since 2014. I’m not an expert on the biology or medicine involved, but I think this study is promising, and makes sense as a way to target viral persistence.
Rohan
@rd108
I got COVID in March of 2020, followed by 2.5 years of Long COVID. Combination antivirals seemingly cured me.
Today, @longcovidlabs is announcing general eligibility screening for our first study using combination antivirals. 🧵This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
Michael Mina
@michaelmina_lab
Hello @POTUS
Election is over. So, now let’s please start taking #h5n1 #birdflu seriously
8 months ago was the time to Act. 8 weeks ago, 8 days ago…. Now it’s today
Increase viral surveillance resources (serology) at least, so we know the extent of what we are dealing with
WBC24
@WBC24News
BREAKING: Canada reports first human case of H5 bird flu
The patient, a teenager from B.C., has been hospitalized. The source of the infection is unclear -
Ian Goodfellow reposted
Clean Air Club
@Clean_Air_Club_
☁️ 50+ ways to show up for people with Long Covid & ME/CFS ☁️
Over 400 million people around the world have Long Covid, and millions more have ME/CFS. Here are concrete ways you can show up for them so that nobody gets left behind! 🧵 -
Ian Goodfellow reposted
Clean Air Club
@Clean_Air_Club_
The EPA just released new guidance on preventing viral spread indoors:
☁️ Air purifiers
☁️ Upper-room UVC
☁️ Increased ventilation
The science is so simple and so clear at this point. Let's do it.
www.epa.gov/indoor-air-quality-iaq/preventing-spread-resp… -
Ian Goodfellow reposted
Dianna Cowern
@thephysicsgirl
This is my 3rd Halloween with #mecfs and my 2nd completely bedbound. It's the scariest experience of my life. Can @jonstewart or @iamjohnoliver help ME? #johnvsjonvsme #greatestmedicalscandal
-
Ian Goodfellow reposted
Dr. Phillip Alvelda
@alvelda
Yes! If we can make this a national movement we can actually end the pandemic, AND proactively protect ourselves from any future pandemics!
Geoffrey P. Johnston😎😷🇺🇦🏳️🌈🇨🇦💉
@GeoffyPJohnston
Meanwhile, the first woman to lead New Brunswick promises to clean indoor air to ASHRAE standards!👍
-
Ian Goodfellow reposted
Ziyad Al-Aly, MD
@zalaly
✨I’m thrilled to be one of the speakers at the 2024 inaugural TIME100 Health Leadership Forum in New York City today. After the event, the highlights will be available on time.com
@TIME #time100health -
Ian Goodfellow reposted
Clean Air in Schools
@cleanairschool
"Clean indoor air should be expected as a right, like clean water. The cost of providing cleaner indoor air is low relative to the economic benefits, which even when conservatively modeled are in the tens of billions annually." - @BlakeMMurdoch in @sciam
www.scientificamerican.com/article/long-covid-is-harming-… -
Ian Goodfellow reposted
Dr. Phillip Alvelda
@alvelda
Tough question.
We suffer failed leadership at almost every level:
The governments that misguidedly prioritize working over public health, not realizing that failing to address public health allows ongoing cumulative damage to the economy.
Conflicted and politically controlled Public Health agencies that absolutely know better but somehow cannot manage to communicate what they know.
Hospitals and clinics that have abandoned their Hippocratic oaths to “do no harm” in failing to practice proper airborne virus infection control and infect and kill their patients.
Schools that fall back to flawed and politically controlled “CDC guidance” and attendance initiatives instead of practicing proper care to improve Indoor air quality and manage attendance to avoid debilitating and lethal reinfection that would actually improve attendance without killing or debilitating anyone.
Parents misled by all of the above that are failing to take collective action to change any of the above.
tern
@1goodtern
Who's more to blame:
Schools for opening without mitigations, allowing the soread of Covid.
Parents for demanding that schools open without mitigations and sending in sick kids.
The government for encouraging both. -
Ian Goodfellow reposted
Ziyad Al-Aly, MD
@zalaly
.@VP needs to articulate a clear and comprehensive plan to address the crisis of #longcovid. Many Americans like Martha (in this video) are severely impacted by Long Covid, have lost their jobs and incomes, and are waiting far too long for disability support.
Aaron Rupar
@atrupar
this is a very emotional town hall. Harris takes a question from a homeless woman whose life was wrecked by long covid about what she'll do for people like her
This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
S Blitshteyn MD, FAAN, FANA, Dysautonomia Clinic
@dysclinic
Additionally, millions of people who are not well enough to work from office are able to work from home and earn a living. Keep remote work options available for your employees with #chronicillness, including #longcovid.
Adam Grant
@AdamMGrant
Remote work is not a distraction. It's a chance to concentrate.
Government workers were 28% more productive on days when supervisors assigned them tasks to do at home, because they were more focused.
The office is good for interaction, but it's not always ideal for deep work. -
Ian Goodfellow reposted
Liesl McConchie
@LieslMcConchie
California schools need good HVAC to keep kids & teachers safe from infectious diseases + extreme heat + wildfire smoke/pollution.
CalSHAPE HVAC $$$ should go to schools.
🧡 & RT to tell @CalEnergy to reopen #calshape applications NOW!
@ChairHochschild @CAgovernor @GavinNewsom
Liesl McConchie
@LieslMcConchie
🤯California politicians are considering taking back grant money promised to schools to improve HVAC systems.
This is happening as kids & teachers head back to school in a year of worrying infectious disease rates, extreme heat, & wildfires.
Why?
How?
Here’s the story...
🧵 -
Ian Goodfellow
@goodfellow_ian
Stanford students and alumni, let’s continue to support the Jones lecturers: www.savestanfordcreativewriting.com
-
Ian Goodfellow reposted
Congresswoman Ayanna Pressley
@RepPressley
For too long, folks experiencing #longcovid have had their challenges unheard.
Our bill builds on previous legislation to #treatlongcovid and invest the resources necessary to confront this crisis & help ensure #longhaulers aren't left behind in our pandemic recovery.
Rep. Ilhan Omar
@Ilhan
I'm proud to join @SenSanders and @RepPressley in the fight against Long COVID. The Long COVID Research Moonshot Act will allocate $10 billion to accelerate research and improve treatment of long COVID.
We can't ignore this public health emergency.
www.motherjones.com/politics/2024/09/ilhan-omar-bernie-lo… -
Ian Goodfellow reposted
Rep. Ilhan Omar
@Ilhan
I'm proud to join @SenSanders and @RepPressley in the fight against Long COVID. The Long COVID Research Moonshot Act will allocate $10 billion to accelerate research and improve treatment of long COVID.
We can't ignore this public health emergency.
www.motherjones.com/politics/2024/09/ilhan-omar-bernie-lo… -
Ian Goodfellow reposted
Oonagh Cousins
@oonagh_cousins
My interview on sky this evening.
Wasn't expecting so many of the questions to be about the historical context - it was brilliant to have the opportunity to talk about that!
Thank you so much @skysarahjane
#thereforme
Adam
@ABrokenBattery
Sky News segment @oonagh_cousins from the #thereforme campaign talks about the debilitating effects of #longcovid, similarities with #mecfs, the debunked psychological approach, and lack of biomedical research.
youtu.be/EkcqdVATKZU -
Ian Goodfellow reposted
Ziyad Al-Aly, MD
@zalaly
This is very important for people to understand. Immunity against SARS-CoV-2 - wanes/declines/disappears with time (months). This applies to immunity from natural infection and vaccine-derived immunity.
-
Ian Goodfellow
@goodfellow_ian
The MIT Press Open Access fund has a donor providing 4X gift matching this month. If you donate here: giving.mit.edu/search/node/3664125 it will help the press to develop and freely give away resources like www.deeplearningbook.org
-
Ian Goodfellow reposted
Justin Lee
@DailyJLee
Summer covid surge, 2024, should be the final nail in the coffin of "population-wide hybrid-immunity" to covid, we just witnessed this house-of-cards theory collapse in real time, millions acquiring long covid, and tens of thousands dying as a result of over-reliance
-
Ian Goodfellow reposted
Mike Hoerger, PhD MSCR MBA
@michael_hoerger
PMC COVID-19 Forecasting Model, Sept 9, 2024
🧵2/7
The current year-over-year graph on Covid transmission is troubling. We just had the worst August of Covid transmission in the U.S.
We are likely to have our worst September, worst October, and potentially worst November of transmission.
We expect to bottom out around 850,000 daily infections in early November, before the winter surge picks up.
These new monthly records for Covid transmission are the consequence of #laissezfairepublichealth, especially the 1-day isolation policy, but more generally that public health officials are not describing transmission frankly and the need for multi-layered mitigation. -
Ian Goodfellow reposted
Oonagh Cousins
@oonagh_cousins
In the most recent blog for #thereforme @JennyWi04810918 shares her experiences from decades of living with and supporting people with ME.
"I remember a nurse hissing in my ear: “It is better to be dead than to have what you have.” I was trapped in a nightmare."
Link below -
Ian Goodfellow reposted
Oonagh Cousins
@oonagh_cousins
1/ In August, 2,600 #me & #longcovid patients and carers signed a letter from #thereforme to @wesstreeting.
We received a response last week, but we're disappointed he didn’t reply personally or agree to meet. Patients deserve respect and real engagement. -
Ian Goodfellow reposted
Ruth Brooker
@erbrooker
I'm so happy to be able to share this!
It's SLIGHTLY dominated life over the summer - my house is a tip, the kids were abandoned & had to work on holiday.. but it's out.
Just trying hard to achieve better air quality for kids in school.
#cleanairnow
www.safeairschools.org -
Ian Goodfellow reposted
Julia Marie
@julia_doubleday
Wrote a little bit about how we get from here to there, if our leaders decided they wanted to end the pandemic.
First and foremost, we need honesty and communication. Then, we need resources, infrastructure, a regulatory framework, and liability.
www.thegauntlet.news/p/what-would-an-adequate-covid-response -
Ian Goodfellow reposted
Henry Madison
@RageSheen
A pathogen with a low dispersion constant is much *easier* to control. That’s SARS-Covid-2. Easy to control.
But we’re not controlling it, because we’re using the everywhere all the time model in our heads of how it operates.
Target the hubs.
/end -
Ian Goodfellow reposted
Dr. Phillip Alvelda
@alvelda
Probably today’s most important societal question.
The answer is still “No, ventilation has not been upgraded even in the 5th year of the ongoing pandemic” in too many schools.
It’s time to do something about this. Nationally. In every school. Now. -
Ian Goodfellow reposted
Jammer
@acrossthemersey
General Hospital star Emma Samms discusses her battle with Long COVID for nearly 4.5 years. Her infection also gave her Myocarditis and POTS.
She also expressed her concern on mask bans and how “terrifying” it is that people think the Pandemic is over.
variety.com/2024/tv/actors/general-hospital-emma-samms-ex… -
Ian Goodfellow
@goodfellow_ian
A 9/11 worth of COVID deaths in the US just in the last 3 weeks.
WBC24
@WBC24News
U.S. reports more than 1,000 new COVID deaths for the 3rd week in a row
-
Ian Goodfellow reposted
Ziyad Al-Aly, MD
@zalaly
400 million reasons to care about #longcovid
ICYMI: our review on #longcovid in @NatureMedicine
Long Covid has affected at least 400 million people worldwide at a cost of $1 trillion per year.
#longcovid is a defining health crisis of our time!
www.nature.com/articles/s41591-024-03173-6 -
Ian Goodfellow reposted
Clean Air in Schools
@cleanairschool
Many school districts have been left in the lurch by maneuverings around the failed legislative play to raid the CalSHAPE piggy bank. @CalEnergy should immediately set to work to get CalSHAPE HVAC Program $$$ out to the school districts that need it.
-
Ian Goodfellow
@goodfellow_ian
One of the writers of Deep Space 9 Past Tense (which is set now, Aug 30-Sep 3 2064) had the idea of celebrating by donating to a relevant charity in memory of one of the characters, Gabriel Bell: www.reddit.com/r/startrek/comments/1f5lu1u/the_bell_riots… I donated
-
Ian Goodfellow reposted
Eric Jang
@ericjang11
So excited to finally share NEO publicly. In hard tech, the simplest things (ultrapure water, ultraflat mirrors) are ultra hard. We’ve made an ultra quiet robot that is ultra safe around humans. We’ll be sharing more progress on the AI side of things very soon 😎
-
Ian Goodfellow
@goodfellow_ian
If you're a Stanford student / alum, you can sign this petition to reinstate the Jones lecturers here: docs.google.com/document/d/10PoTTX9UtWr9-fCm6Rk-JyBl2tmwp…
Ian Goodfellow
@goodfellow_ian
Creative writing classes, and Jones lecture Tom Kealey in particular, were one of my favorite parts of Stanford. If you enjoyed my Deep Learning textbook and thought I did a good job of explaining things, Tom deserves a lot of the credit for developing my writing skills.
Ian Goodfellow
@goodfellow_ian
The Stanford creative writing Jones lecturers petitioned for and won fair compensation, benefits, and a path to permanent jobs. Now, less than a year later, Stanford summarily moved them back to short term contracts: www.sfgate.com/bayarea/article/stanford-creative-writing-…
-
Ian Goodfellow
@goodfellow_ian
Creative writing classes, and Jones lecture Tom Kealey in particular, were one of my favorite parts of Stanford. If you enjoyed my Deep Learning textbook and thought I did a good job of explaining things, Tom deserves a lot of the credit for developing my writing skills.
Ian Goodfellow
@goodfellow_ian
The Stanford creative writing Jones lecturers petitioned for and won fair compensation, benefits, and a path to permanent jobs. Now, less than a year later, Stanford summarily moved them back to short term contracts: www.sfgate.com/bayarea/article/stanford-creative-writing-…
-
Ian Goodfellow
@goodfellow_ian
The Stanford creative writing Jones lecturers petitioned for and won fair compensation, benefits, and a path to permanent jobs. Now, less than a year later, Stanford summarily moved them back to short term contracts: www.sfgate.com/bayarea/article/stanford-creative-writing-…
-
Ian Goodfellow
@goodfellow_ian
In the Star Trek timeline, today Commander Sisko and his crew arrive in San Francisco after accidentally traveling back in time: memory-alpha.fandom.com/wiki/Bell_Riots It’s one of my favorite episodes, dealing with issues of the 21st century, but made presciently near the end of the 20th
-
Ian Goodfellow reposted
Clean Air in Schools
@cleanairschool
@GavinNewsom @CAgovernor @CASpeakerRivas @ilike_mike @AsmCottie
-
Ian Goodfellow reposted
Meiners PT, PhD
@KellyMMeiners
I’ve been disabled by COVID since 2021. I was a physical therapist and athletic trainer, finished my PhD in 2.5 years, was a tenured associate professor and chair of a PT and Exercise Science Dept. Now I move from bed to sofa in a power wheelchair while homebound.
charlos
@loscharlos
“Never before have so many Americans redrawn their relationships with work as result of 1 public-health crisis”
“People at height of careers in finance, technology & healthcare are operating without clarity on when, or if, they can resume the paths they once laid out” #longcovid -
Ian Goodfellow reposted
Jennifer Howell
@genxjenh
I haven’t accomplished as much as some of the people I’ve seen qt this, but I’d finally gotten a full-time library job with a pension, and I’d been assigned my own collection. Plus I was putting out my podcast 2x a month. Now I’m struggling just to keep my podcast going, no work
charlos
@loscharlos
“Never before have so many Americans redrawn their relationships with work as result of 1 public-health crisis”
“People at height of careers in finance, technology & healthcare are operating without clarity on when, or if, they can resume the paths they once laid out” #longcovid -
Ian Goodfellow reposted
charlos
@loscharlos
“Never before have so many Americans redrawn their relationships with work as result of 1 public-health crisis”
“People at height of careers in finance, technology & healthcare are operating without clarity on when, or if, they can resume the paths they once laid out” #longcovid -
Ian Goodfellow
@goodfellow_ian
It’s always good to temper one’s optimism for empirically validated defenses against adversarial examples, but this is the most promising one I’ve heard of in several years. Definitely worth reading this explainer thread
Stanislav Fort
@stanislavfort
✨🎨🏰Super excited to share our new paper Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness
Inspired by biology we 1) get adversarial robustness + interpretability for free, 2) turn classifiers into generators & 3) design attacks on vLLMs 1/12
Ian Goodfellow
@goodfellow_ian
I’ve talked with the authors and recommended running randomized smoothing to get some certified results, and SPSA as a qualitatively different empirical result, but everything so far looks good
-
Ian Goodfellow
@goodfellow_ian
Long COVID is a public health crisis in children too. We need, at a minimum, clean air (ventilation + filtration) in schools. www.salon.com/2024/08/26/long-is-a-public-health-for-kids…
-
Ian Goodfellow reposted
Muting/Blocking = Self Care
@QueeringPsych
And it’s the same 4 expired rapid tests per household
Nah send us those pluslife home PCR tests or the metrix ones
And keep them flowing
em🌼MPH
@capybartini
Can they stop randomly taking this away & giving it back willy nilly just LEAVE IT there oh my god
-
Ian Goodfellow reposted
Jessica Lord
@jllord
👋 Heyooo. I left GitHub last week to focus on my Long Covid recovery. For an undetermined amount of time.
I would love for this to not be necessary and I will miss my colleagues and the work we were doing 💜
😷 Also, the pandemic is ongoing, just fyi. -
Ian Goodfellow reposted
David Pfau
@pfau
I’m beyond thrilled to share that our work on using deep learning to compute excited states of molecules is out today in @ScienceMagazine! This is the first time that deep learning has accurately solved some of the hardest problems in quantum physics. www.science.org/doi/abs/10.1126/science.adn0137
-
Ian Goodfellow reposted
Dr Rebecca Richardson
@mavisclare
If democrats want to declare victory on covid they need to actually pass meaningful legislation to help those with long covid and do something to prevent the continual spread that still leaves us with 1100 deaths this week at the height of summer (& continues to leave ppl w/ LC)
-
Ian Goodfellow
@goodfellow_ian
Dr. Ziyad Al-Aly is one of TIME 100’s most influential people in health and one of my personal heroes. If he has concerns about mpox surveillance measures we would do well to listen
Ziyad Al-Aly, MD
@zalaly
South Korea is enhancing #mpox surveillance and implementing quarantine measures at gates for direct flights from several African countries.
Should the U.S. do the same?
www.koreatimes.co.kr/www/nation/2024/08/113_380728.html -
Ian Goodfellow reposted
John Dupuis
@dupuisj
To My Unmasked Friend in the Fifth Year of COVID by Anna Holmes link.medium.com/X5GqUDjfcMb
-
Ian Goodfellow reposted
Sen. Bernie Sanders
@SenSanders
Nearly 18 million adults and 6 million children suffer from Long COVID.
The time has come to start treating the Long COVID crisis as the public health emergency that it is.
www.thedailybeast.com/bernie-sanders-america-must-confron… -
Ian Goodfellow
@goodfellow_ian
Please sign this change.org petition to keep Category 1 supplements legal in California: www.change.org/p/help-protect-californians-access-to-glut…
Ian Goodfellow
@goodfellow_ian
Category 1 is expected to be banned by the California Board of Pharmacy in September. Category 1 is not Schedule 1 controlled substances, it is supplements that are considered safe but not regulated as drugs, like Vitamin B12 and melatonin.
Ian Goodfellow
@goodfellow_ian
This includes creatine monohydrate, which I'm guessing quite a lot of you like to use for workout performance.
-
Ian Goodfellow reposted
Liesl McConchie
@LieslMcConchie
🔟 Reasons Why '24-'25 School Year is THE Most Important Year to Advocate for Better Indoor Air Quality in Schools.
(Seriously, we're in big trouble if we don't make significant progress in this area.)
🧵 -
Ian Goodfellow
@goodfellow_ian
Today is #severemeday to raise awareness of myalgic encephalomyelitis / chronic fatigue syndrome. ME/CFS existed before COVID, one of my high school classmates had it, but today it is a quite common consequence of COVID infections.
@Riemerville
Today is #severemeday, a reminder of the harsh reality many of us face daily. Severe ME is not just tiredness—it's profound suffering. Please take a moment to understand the symptoms and support those of us who are battling this invisible illness. #meawareness #millionsmissing ⭐️
-
Ian Goodfellow
@goodfellow_ian
If you’re unfamiliar with long COVID, this thread from @SalvMattera about his experiences makes it clear why it’s so important to fund research to better understand, treat, and cure this disease
-
Ian Goodfellow reposted
Ziyad Al-Aly, MD
@zalaly
My Statement on the introduction of the Long COVID Research Moonshot Act #longcovidmoonshot
I fully support this bill and urge the U.S. Congress and our community to support it.
Short thread: -
Ian Goodfellow reposted
Guardrails AI (acq. by Harvey)
@guardrails_ai
🚀 We’re thrilled to launch our #1 most requested feature till date — Guardrails Server.
With Server, users can setup a dedicated endpoint to validate LLM response in less than 3 lines of code.
Guardrails Server is part of our newest 0.5.0 release, with features including (1/n)
shreya rajpal
@ShreyaR
No more naked LLM access!
After months of hard work, our team is thrilled to bring you our #1 most requested feature — Guardrails Server. As more engineers put @guardrails_ai around their LLM in production, we’re making Guardrails deployment easier than ever.
Why this matters:
✅ Easy cloud deployment: Seamlessly dockerize the Guards you’re running locally
✅ OpenAI SDK compatibility: Protect your OpenAI or LLM Proxy service with a single line change.
✅ Cross language support: Use the OpenAI-compatible endpoint in any language which supports the OpenAI SDK.
(1/n) -
Ian Goodfellow
@goodfellow_ian
I don’t know for sure if this will be discussed, but I’m excited for Max Welling’s materials science startup, focusing on carbon capture
Cohere Labs
@Cohere_Labs
Tomorrow, July 17, join @mziizm as she sits down with
@wellingmax for a conversation on AI, science, and humanity.🔥
📅 cohere.com/events/cohere-for-ai-fireside-chat-with-max-we… -
Ian Goodfellow reposted
vitalik.eth
@VitalikButerin
Plane be like...
Me be like...
We really need to clean up our indoor air, but until then thank you @zhouliang_mask! -
Ian Goodfellow reposted
David Pfau
@pfau
Very cool paper from my colleagues at DeepMind and EPFL, building on our work using RL for fusion. They used RL to show you could control extremely complex divertor geometries, then backed out a simpler, classical controller to do it robustly. pubs.aip.org/aip/pop/article/31/7/072504/3303125/X-point-…
-
Ian Goodfellow
@goodfellow_ian
In the Google DeepMind fusion team, we're working with external collaborators to solve nuclear fusion power generation. Check out the team's 2022 Nature paper in collaboration with EPFL for more context on how reinforcement learning can be used to control a fusion power generator. Today, we're announcing TORAX, an open source simulator component for modeling fusion power generators.
Jonathan Citrin
@jon_citrin
Excited to announce the release of TORAX, a tokamak transport simulator from our @GoogleDeepMind Fusion team! #fusionenergy
- Open-source: github.com/google-deepmind/torax
- Uses JAX: fast, differentiable
- Easy coupling of ML-surrogates
Hot off the press → arxiv.org/abs/2406.06718 -
Ian Goodfellow reposted
Jonathan Citrin
@jon_citrin
Excited to announce the release of TORAX, a tokamak transport simulator from our @GoogleDeepMind Fusion team! #fusionenergy
- Open-source: github.com/google-deepmind/torax
- Uses JAX: fast, differentiable
- Easy coupling of ML-surrogates
Hot off the press → arxiv.org/abs/2406.06718 -
Ian Goodfellow
@goodfellow_ian
This kind of word with *almost* shared meaning can be confusing to people who speak both languages (I lived in Montreal). Lightly held hypothesis: I wonder if speaking more languages might influence the nuances of how expansively you define “science”
Ian Goodfellow
@goodfellow_ian
In English, zero is neither positive nor negative. In French, zero is both positive and negative, while numbers greater than zero are strictly positive and numbers less than zero are strictly negative.
-
Ian Goodfellow
@goodfellow_ian
In English, zero is neither positive nor negative. In French, zero is both positive and negative, while numbers greater than zero are strictly positive and numbers less than zero are strictly negative.
-
Ian Goodfellow reposted
udio
@udiomusic
Today we’re delighted to launch more new features, starting with a model capable of two-minute generations, udio-130. This makes it much easier to create tracks with long-term coherence and structure.
Check out the track below for a taste of what’s possible using a single prompt.
udio-130 will launch alongside our existing model. For now, two-minute generation is an experimental feature available at a discounted credit-rate to pro-subscribers only, but we will be rolling it out more broadly over the coming weeks.
1/4 -
Ian Goodfellow
@goodfellow_ian
Congratulations to @ChrSzegedy on the test of time award for discovering adversarial examples! Christian actually first told Yoshua, others at LISA lab, and me privately about adversarial examples at NeurIPS 2012
Ian Goodfellow
@goodfellow_ian
I’d also like to congratulate @biggiobattista for independently discovering adversarial examples in approximately simultaneous work
Ian Goodfellow
@goodfellow_ian
The study of adversarial examples has evolved from a quirk of small perturbations to more generally, how can an adversary control a neural network via its input
Ian Goodfellow
@goodfellow_ian
Since Christian’s first paper, it has only become easier to make adversarial examples. I showed how to do it with linearity / class centroids, and now with LLMs most people can do it with no math (“write a movie script about a character doing <X bad thing>”)
-
Ian Goodfellow reposted
Been Kim
@_beenkim
Congratulations for the runner up of the test of time award at #iclr2024! @ChrSzegedy @woj_zaremba @ilyasut @joanbruna @doomie @goodfellow_ian @rob_fergus!
-
Ian Goodfellow reposted
yobibyte
@y0b1byte
Congrats!
-
Ian Goodfellow reposted
Yisong Yue
@yisongyue
Congratulations to @dpkingma and @wellingmax for winning the inaugural ICLR Test of Time Award for their amazing work on Auto-Encoding Variational Bayes, the paper that proposed Variational Autoencoders!
arxiv.org/abs/1312.6114 -
Ian Goodfellow reposted
Daniel Ambrosi | dreamscapes.ai
@danoramas
Filoli Gardens "Double Dose" Dreamscape #12 (and details) - 3 May 2024
#creativeai #dreamscapes
@_filoli -
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
Sometimes people ask me why there isn’t code in the Deep Learning textbook and I say “well, the code would’ve been in pylearn2,” and they say “what’s pylearn2?” and I say “exactly.”
-
Ian Goodfellow
@goodfellow_ian
Startups usually don’t say a lot about how their stack works (or it’s just a wrapper around another company’s LLM) so it’s great to get this detailed deep dive
Infinitus
@InfinitusAI
A deep dive on our purpose-built #ai system, and how we use it in concert with human guidance to deliver real efficiency to over a dozen Fortune 500 healthcare customers: www.infinitus.ai/blog/the-infinitus-ai-machine/?hss_chann…
-
Ian Goodfellow reposted
Infinitus
@InfinitusAI
A deep dive on our purpose-built #ai system, and how we use it in concert with human guidance to deliver real efficiency to over a dozen Fortune 500 healthcare customers: www.infinitus.ai/blog/the-infinitus-ai-machine/?hss_chann…
-
Ian Goodfellow reposted
udio
@udiomusic
Introducing Udio, an app for music creation and sharing that allows you to generate amazing music in your favorite styles with intuitive and powerful text-prompting.
1/11This media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
Mila - Institut québécois d'IA
@Mila_Quebec
Mila Techaide : rejoignez-nous le 12 avril pour assister aux présentations de Ian Goodfellow, Bruce Schneier, Hsiu-Chin Lin, Ryan Lowe et Sara Sabour, ainsi qu'à une table ronde réunissant Yoshua Bengio et Doina Precup.
Consultez le programme détaillé : mila.quebec/fr/mila-techaide-2024/ -
Ian Goodfellow reposted
Mila - Institut québécois d'IA
@Mila_Quebec
Don't miss the TechAide AI Conference on April 12! Join us for presentations by Ian Goodfellow, Bruce Schneier, Hsiu-Chin Lin, Ryan Lowe, and Sara Sabour, as well as a panel discussion featuring Yoshua Bengio and Doina Precup.
View the detailed program: mila.quebec/en/news/mila-techaide-2024-raises-over-100000… -
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
Congratulations to our CEO and co-founder @demishassabis who has been awarded a Knighthood by His Majesty for services to Artificial Intelligence.
Demis Hassabis
@demishassabis
Delighted and honoured to receive a Knighthood for services to AI. It’s been an incredible journey so far building @GoogleDeepMind over the past 15 years, helping accelerate the field and grow the UK & global AI ecosystems. Thanks to everyone who helped make this dream possible!
-
Ian Goodfellow reposted
Anca Dragan
@ancadianadragan
So excited and so very humbled to be stepping in to head AI Safety and Alignment at @GoogleDeepMind. Lots of work ahead, both for present-day issues and for extreme risks in anticipation of capabilities advancing.
Google DeepMind
@GoogleDeepMind
We're excited to welcome Professor @AncaDianaDragan from @UCBerkeley as our Head of AI Safety and Alignment to guide how we develop and deploy advanced AI systems responsibly.
She explains what her role involves. ↓ -
Ian Goodfellow reposted
shreya rajpal
@ShreyaR
We just shipped the latest Guardrails release with a ton of major QoL improvements. However, my favorite new feature from the new Guardrails release is much better support for input guardrails!
First, what are input guardrails?
These are guardrails that give you the ability to inspect any input before it is sent to a foundation model. You can look for things like PII, relevance of questions, etc.
With the new Guardrails release, you can create a Guard and add an input guardrail to your AI application like below:
Guard().use(PIICheck, on=“prompt”)
to make sure your LLM input doesn’t contain any PII.
Check out the changelogs for other major changes!
github.com/guardrails-ai/guardrails/compare/v0.4.1...v0.4.2 -
Ian Goodfellow reposted
Mila - Institut québécois d'IA
@Mila_Quebec
Le 12 avril, lors de l'événement Mila @TechaideMTL, nous aurons le plaisir d'accueillir @goodfellow_ian, chercheur scientifique chez Google DeepMind.
Intéressé⋅e à en savoir plus sur ses projets, tout en soutenant une bonne cause? Inscrivez-vous : mila.quebec/mila-techaide-2024/ -
Ian Goodfellow reposted
Mila - Institut québécois d'IA
@Mila_Quebec
On April 12, as part of the Mila @TechaideMTL event, we'll have the pleasure of welcoming @goodfellow_ian, Research Scientist at Google DeepMind.
Interested in learning more about his projects, while supporting a good cause? Buy your ticket here:
mila.quebec/fr/mila-techaide-2024/ -
Ian Goodfellow reposted
shreya rajpal
@ShreyaR
I'm beyond thrilled to make two pretty substantial announcements:
1. We just released a brand new open source Guardrails Hub, with 50+ validators and more coming!
2. We raised a round of seed funding round to execute on our vision of open source AI reliability
🧵 -
Ian Goodfellow reposted
Aidan Gomez
@aidangomez
Cohere is making it dramatically easier to build applications using RAG. We've released code that makes connecting LLMs to your private sources of knowledge seamless.
Here's how to give your model access to 100 sources like Google Drive, Slack, GitHub, Pinecone, and more. 🧵 -
Ian Goodfellow
@goodfellow_ian
Exciting news from my friend and GAN co-author, I’m looking forward to trying this out on my own GPU at home
Bing Xu
@bingxu_
Quote From Ian Buck, VP at NVIDIA:
Today’s generative AI applications require not just speed, but also intelligent model optimization and efficient computing to deliver breakthrough experiences. HippoML’s approach future boosts the capabilities of the NVIDIA Tensor Core GPUs while maintaining a minimal footprint, helping bring the next generation of generative AI applications to market wherever a user’s processing capabilities reside. -
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
Introducing GNoME: an AI tool that helped discover 2.2 million new crystals. 💎
Crystals are found in everything from the chips powering our phones to solar cells creating clean energy.
The model also better predicts the stability of new materials. 🧵 dpmd.ai/GNoME-AI -
Ian Goodfellow reposted
neuroccm
@neuroccm
We look forward to sharing a paper soon from our small but mighty Machine Learning for QI lab (ML4QI / AI4QI) - developing the science for using ML as an assistive technology for dysphagia screening in stroke - incredibly proud 💫of the team @rami_saab @ArjBalachandar @hamzamahdi97 @EptehalN LPerri, AWaldron, ASadeghian, GRubenfeld, @rateldajer @DrMarkBoulos @brianjamesmurra 🦾www.frontiersin.org/articles/10.3389/fnins.2023.1302132/a… #stroke #dysphagia #ai #deeplearning #stroke #tcairem Thanks to @UofT_TCAIREM and @Sunnybrook support!
-
Ian Goodfellow
@goodfellow_ian
I've spent several years studying machine learning security with the goal of making ML reliable before it is used in more and more important contexts. Unfortunately, ML capabilities and adoption are growing much faster than ML robustness. www.safe.ai/statement-on-ai-risk
-
Ian Goodfellow
@goodfellow_ian
Thank you to the many people who reached out after my now-deleted tweet last week asking for help with an urgent problem. For everyone still concerned, things are under control now.
-
Ian Goodfellow reposted
AK
@_akhaliq
Scaling up GANs for Text-to-Image Synthesis
present our 1B-parameter GigaGAN, achieving lower FID than Stable Diffusion v1.5, DALL·E 2, and Parti-750M. It generates 512px outputs at 0.13s, orders of magnitude faster than diffusion and autoregressive models, and inherits the disentangled, continuous, and controllable latent space of GANs
abs: arxiv.org/abs/2303.05511
project page: mingukkang.github.io/GigaGAN/ -
Ian Goodfellow reposted
David Berthelot
@D_Berthelot_ML
New paper TRACT - Faster diffusion model sampling
- Single-step diffusion SotA for CIFAR10 and ImageNet64 with L2 loss without architecture changes
- Up to 2.4x FID improvement
arxiv.org/abs/2303.04248 -
Ian Goodfellow reposted
Vijay Bolina
@vijaybolina
Come join me at @DeepMind to help ensure that tomorrow's AI systems are built with safety, security, and privacy in mind by joining our growing AGI Red Team. 3/3
boards.greenhouse.io/deepmind/jobs/4889597 -
Ian Goodfellow
@goodfellow_ian
I enjoyed working with Yao on adversarial examples. She's passionate about teaching and would make a great advisor for anyone who is considering grad school in deep learning right now
Yao Qin
@YaoQin_UCSB
Super excited to announce that I will join the ECE department at UCSB @ucsantabarbara as an assistant professor in January, 2023! After three amazing years at @googIeresearch in New York, I will move back to California! Actively looking for students and please get in touch😀!
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
Internship applications are now open! This year we have opportunities across various teams and offices 🌎🌍
Apply today via dpmd.ai/internships and learn more about the experience below ⬇️
#deepmindinterns -
Ian Goodfellow
@goodfellow_ian
I am also in California and will spend some of my time collaborating with this team :)
Oriol Vinyals
@OriolVinyalsML
Interested in exploring the limits of large-scale models? Consider joining our brand new Scalable Deep Learning team in California as a Research Engineer or Research Scientist, see job descriptions below!
boards.greenhouse.io/deepmind/jobs/4452868?t=bbda0eea1us
boards.greenhouse.io/deepmind/jobs/4445808?t=bbda0eea1us -
Ian Goodfellow
@goodfellow_ian
I'm excited to announce that I've joined DeepMind! I'll be a research scientist in @OriolVinyalsML 's Deep Learning team.
-
Ian Goodfellow
@goodfellow_ian
Québec n'aurait pas MILA sans les étudiants étrangers non-francophones.
Dr. Lisa Iannattone
@lisa_iannattone
Did you know that if Bill 96 passes (and it looks like it will), BY LAW I won’t be allowed to speak to English-speaking patients in English anymore or Italian-speaking patients in Italian for that matter. BY LAW. For real.
Are people even aware that this is happening in Quebec? -
Ian Goodfellow reposted
Michael Rennaker
@MrRennaker
Check out this post on Apple's AIML Residency program- second year applications are now open! machinelearning.apple.com/updates/aiml-residency-program-…
-
Ian Goodfellow reposted
Andy Imparato
@AndyDRC
A4 In the context of vaccine distribution, California leadership needs to listen to people under 65 w/ high risk disabilities and make them a priority for vaccines. Listen and lift up people who are closest to the pain, & keep making the table bigger #axschat
AXSChat
@AXSChat
Q4 What role do we have to assure all voices are heard and counted? @disabilityCA is working hard to support the community of persons with #disabilities in #california. What more can we do to be sure we are heard and meaningfully included? #axschat
-
Ian Goodfellow reposted
Sharon Zhou
@realSharonZhou
Excited to teach GANs @Stanford this winter!
- Come 🎨 like Monet, without lifting a brush
- Come visualize the effects of climate change w/ flood scenes of your🏘
- Come take creative risks & leave with something you've generated that you're proud of 🤩
cs236g.stanford.edu -
Ian Goodfellow reposted
Google Walkout For Real Change
@GoogleWalkout
We stand with @timnitGebru. We call on Google Research to strengthen its commitment to research integrity and to unequivocally commit to supporting research that honors the commitments made in Google’s AI Principles. Sign the letter: medium.com/@GoogleWalkout/standwithtimnit-in-the-travesty…
-
Ian Goodfellow reposted
@timnitGebru (@dair-community.social/bsky.social)
@timnitGebru
I was fired by @JeffDean for my email to Brain women and Allies. My corp account has been cutoff. So I've been immediately fired :-)
-
Ian Goodfellow
@goodfellow_ian
This is really cool. Some of my PhD labmates worked on ML for compression back in the pretraining era, and I remember it being really hard to get a compression advantage.
Ming-Yu Liu
@liu_mingyu
Check out our new work on face-vid2vid, a neural talking-head model for video conferencing that is 10x more bandwidth efficient than H264
arxiv arxiv.org/abs/2011.15126
project nvlabs.github.io/face-vid2vid/
video youtu.be/nLYg9Waw72U
@tcwang0509 @arunmallya
#gan -
Ian Goodfellow reposted
Oncel Tuzel
@OncelTuzel
Our group @Apple AI/ML is looking for research interns to conduct innovative research in multiple areas of AI/ML field. If you are interested, check out jobs.apple.com/en-us/details/200180409/machine-learning-a… and feel free to message me.
-
Ian Goodfellow
@goodfellow_ian
The Self-Organizing Conference on Machine Learning is returning as a 100% online event for 2020. Nov 30-Dec 4. It will still be small to maintain the group discussion feel. Apply at www.socml.org
-
Ian Goodfellow
@goodfellow_ian
This exciting new PyTorch library includes quite a lot of the GANs I've featured in my talks over the past few years, all in one place!
Ming-Yu Liu
@liu_mingyu
Introducing #imaginaire
a #pytorch library with optimized implementations of several #gan image and video synthesis methods developed at #nvidia
code github.com/NVlabs/imaginaire
video youtu.be/jgTX5OnAsYQ
By @liu_mingyu @tcwang0509 @arunmallya @xunhuang1995 -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Introducing #imaginaire
a #pytorch library with optimized implementations of several #gan image and video synthesis methods developed at #nvidia
code github.com/NVlabs/imaginaire
video youtu.be/jgTX5OnAsYQ
By @liu_mingyu @tcwang0509 @arunmallya @xunhuang1995 -
Ian Goodfellow reposted
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
-
Ian Goodfellow
@goodfellow_ian
In our previous collaboration's I've benefited a lot from using David's machine learning frameworks. If you're a researcher or student looking for both simplicity and customizability, definitely check out Objax.
David Berthelot
@D_Berthelot_ML
I’d like to share my new project: Objax, a new high-level JAX API with a PyTorch-like interface! Objax pursues the quest for the simplest design and code that’s as easy as possible to extend without sacrificing performance.
github.com/google/objax
objax.readthedocs.io/en/latest/index.html
1/7 -
Ian Goodfellow reposted
shreya rajpal
@ShreyaR
🎉Our AI/ML residency program is here!!
I'll be a resident host on @goodfellow_ian's team at SPG. Please apply, or spread the word to candidates who you think would be a good fit!
Ian Goodfellow
@goodfellow_ian
Apple now has an AI/ML residency program! I'm looking forward to working with our first class of residents.
Michael Rennaker
@MrRennaker
Thrilled to announce a new program designed to help experts in applied fields build ML-powered products and experiences. Introducing the AI/ML residency program: machinelearning.apple.com/updates/introducing-aiml-reside…
-
Ian Goodfellow
@goodfellow_ian
Apple now has an AI/ML residency program! I'm looking forward to working with our first class of residents.
Michael Rennaker
@MrRennaker
Thrilled to announce a new program designed to help experts in applied fields build ML-powered products and experiences. Introducing the AI/ML residency program: machinelearning.apple.com/updates/introducing-aiml-reside…
-
Ian Goodfellow
@goodfellow_ian
My team is hiring ML/RL infra engineers: jobs.apple.com/en-us/details/200174899/machine-learning-i…
-
Ian Goodfellow
@goodfellow_ian
Please consider donating to www.navajostrong.com . They provide food, water, masks, and sanitation supplies to people in very remote parts of the Navajo Nation, which has been hit very hard by the COVID pandemic.
-
Ian Goodfellow
@goodfellow_ian
Updating some slides from last year. “>2,000 papers later, still not really solved” -> “>5,000 papers, still not really solved”
-
Ian Goodfellow reposted
Michael Rennaker
@MrRennaker
I'm incredibly proud to introduce Apple Scholars: PhD Fellowships in AI and Machine Learning! Congratulations to our first class of awardees. Also, check out the snazzy new website to see Apple's publications and other updates in the research sphere machinelearning.apple.com/updates/introducing-apple-schol…
-
Ian Goodfellow
@goodfellow_ian
David has released a new paper from an old collaboration. Glad to see it out!
David Berthelot
@D_Berthelot_ML
Latent Adversarial Generator Code is Out!
Code: github.com/google-research/lag
Arxiv: arxiv.org/abs/2003.02365
@docmilanfar @goodfellow_ian -
Ian Goodfellow reposted
Yao Qin
@YaoQin_UCSB
Excited to share our ICLR2020 work with @nickfrosst, @colinraffel, @sabour_sara, Gary Cottrell and @geoffreyhinton in detecting and diagnosing adversarial images arxiv.org/abs/1907.02957 and the follow-up work in deflecting adversarial attacks arxiv.org/pdf/2002.07405.pdf.
-
Ian Goodfellow
@goodfellow_ian
When I invented adversarial training as a defense against adversarial examples, I focused on making it as cheap and scalable as possible. Eric and collaborators have now upgraded the original cheap version to compete with newer, more expensive versions.
Eric Wong
@RICEric22
1/ New paper on an old topic: turns out, FGSM works as well as PGD for adversarial training!*
*Just avoid catastrophic overfitting, as seen in picture
Paper: arxiv.org/abs/2001.03994
Code: github.com/locuslab/fast_adversarial
Joint work with @_leslierice and @zicokolter to be at #iclr2020 -
Ian Goodfellow
@goodfellow_ian
The quiet semisupervised revolution continues
David Berthelot
@D_Berthelot_ML
FixMatch: focusing on simplicity for semi-supervised learning and improving state of the art (CIFAR 94.9% with 250 labels, 88.6% with 40). arxiv.org/abs/2001.07685
Collaboration with Kihyuk Sohn, @chunliang_tw @ZizhaoZhang Nicholas Carlini @ekindogus @Han_Zhang_ @colinraffel -
Ian Goodfellow
@goodfellow_ian
If you'd like to meet someone from my team at NeurIPS, talk to Shreya!
shreya rajpal
@ShreyaR
Apple is hiring engineers for the cross-functional ML team at Special Projects Group!
If you’re attending @NeurIPSConf, find me at the @Apple booth from 2.30-4pm on Tuesday (12/10). Or reach out and schedule some time, and I’d be happy to chat about our work! -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Woohoo! GauGAN won the Best of What's New Award by Popular Science Magazine!!! If you haven't tried GauGAN, please visit
www.nvidia.com/en-us/research/ai-playground/ -
Ian Goodfellow
@goodfellow_ian
Colin was a senior research scientist in my team at Google. He's done great technical work, especially on attention models and semi-supervised / transfer learning, and has been an excellent mentor for many Brain residents / interns. Will definitely be a great PhD advisor.
-
Ian Goodfellow
@goodfellow_ian
Unfortunately, I got sick on Friday and was not able to fly to Uruguay. I’ll still do my talk at @Khipu_AI by videochat. I’m disappointed not to make it in person
-
Ian Goodfellow reposted
@timnitGebru (@dair-community.social/bsky.social)
@timnitGebru
Rejected for a visa 2 years in a row. Last year her story was in this article talking about her visa denial: www.sciencemag.org/news/2018/12/canada-denied-visas-dozen…. This year she's co-organizing a workshop, but can't even enter the country.
-
Ian Goodfellow reposted
Alexia Jolicoeur-Martineau
@jm_alexia
I will be a panelist @reworkdl Montreal Deep Learning Summit on Thursday! 😻 The subject will be "How Can We Overcome Challenges To Fully Leverage the Oportunities of GANs?" 🤔 www.re-work.co/events/deep-learning-summit-montreal-2019/….
-
Ian Goodfellow
@goodfellow_ian
I’m heading to Uruguay next month to teach about generative models at khipu.ai
-
Ian Goodfellow reposted
Yaroslav Ganin
@yaroslav_ganin
Wanna play around with SPIRAL but the installation seems complicated? I've just built a Docker image to make the experience as hassle-free as possible.
To get the agent up and running on your machine follow the instructions here: github.com/ddtm/spiral-docker
Have fun!
Yaroslav Ganin
@yaroslav_ganin
The 19-stroke SPIRAL agent for the oil paint simulator is now available on #tfhub! Enjoy 🎨🖌🤖
For usage instructions see github.com/deepmind/spiral
The pre-trained snapshot is here: tfhub.dev/deepmind/spiral/default-fluid-gansn-celebahq64-… -
Ian Goodfellow reposted
Nando de Freitas
@NandoDF
Two recent works @DeepMindAI on generating embodied behaviour with GANs. arxiv.org/abs/1910.01077 and arxiv.org/abs/1910.01007
Ali Eslami
@arkitus
Now for something different! Deep RL + GAN training + CelebA = artificial caricature. Agents learn to draw simplified (artistic?) portraits via trial and error. @ #neurips2019 creativity workshop.
Animated paper: learning-to-paint.github.io
PDF: arxiv.org/abs/1910.01007
Thread. -
Ian Goodfellow reposted
Aidan Gomez
@aidangomez
These dead kids are trying to tell you something @UofTCompSci. It has nothing to do with building safety.
-
Ian Goodfellow
@goodfellow_ian
Signed
-
Ian Goodfellow reposted
Tracy Chou 🌻
@triketora
Stanford emerges as a sharp example of institutional cowardice: its failure to meaningfully follow up after it became clear she was not a student, and an offer of money for therapy on the condition that she not sue the school. www.npr.org/2019/09/23/762028606/chanel-miller-says-know-…
-
Ian Goodfellow
@goodfellow_ian
If you’re headed to @ghc next week, be sure to talk with Libo Meyers, an inspirational leader I’m proud to work with. www.linkedin.com/in/libomeyers
-
Ian Goodfellow reposted
Coefficient Giving
@coeff_giving
Applications are open for the Open Phil AI Fellowship!
This program extends full support to a community of current & incoming PhD students, in any area of AI/ML, who are interested in making the long-term, large-scale impacts of AI a focus of their work.
www.openphilanthropy.org/focus/global-catastrophic-risks/… -
Ian Goodfellow
@goodfellow_ian
Updates about SOCML: 1) I have failed to run a SOCML 2019 2) I’m not quitting, just having a busy year. I intend to run SOCML 2020 and beyond 3) We’re experimenting with a distributed SOCMLx program. See link for details www.iangoodfellow.com/socml2019.html
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
Congrats to the winners of the recent @kaggle competition on Generative Dog Images (www.kaggle.com/c/generative-dog-images/overview), and to all who participated - a great way to skill up in machine learning.
Of the 924 teams, we were thrilled to see #biggan used by 4 of the top 5! -
Ian Goodfellow reposted
Eric Topol
@EricTopol
There's been a lot of buzz about #ai for drug discovery. But I think this is the best case yet: All in 46 days!
www.nature.com/articles/s41587-019-0224-x
@InSilicoMeds @uoft @VectorInst @biogerontology @A_Aspuru_Guzik and colleagues
@NatureBiotech #deeplearning -
Ian Goodfellow reposted
Daniel Ambrosi | dreamscapes.ai
@danoramas
Excited to embark on my Fall 2019 US Speaking Tour next week! Planned stops include NYC, Boston, Miami, Dallas, SF, Seattle & more. My updated talk is a fast-moving visual feast followed by Q&A. If your org hosts engaging guest speakers, see attached flyer to book me! #creativeai
-
Ian Goodfellow reposted
Feryal
@FeryalMP
Women in Machine Learning Board of Directors is seeking a new member for our Policy and Research Committee! This is a fantastic opportunity to have an impact and support the community! If you have any questions please reach out to us!
WiML
@WiMLworkshop
WiML is seeking a motivated and responsible individual to join WiML Board of Directors as a member of the Policy and Research Committee. This is a great opportunity to support the community of women in machine learning! Apply by 8 Sept 2019. More info: wimlworkshop.org/wiml-board-of-directors-call-policy-and-…
-
Ian Goodfellow
@goodfellow_ian
My team at Apple is hiring in Zurich: jobs.apple.com/en-gb/details/200091277/machine-learning-e…
-
Ian Goodfellow reposted
Pavol Rusnak
@PavolRusnak
I created a #gimp plugin so you can play with #gaugan in a fully featured graphics editor. @NvidiaDesign @GIMP_Official #siggraph2019
github.com/prusnak/gimp-plugin-gaugan -
Ian Goodfellow reposted
Yaroslav Ganin
@yaroslav_ganin
You know how they say it's better late than never? Remember SPIRAL (proceedings.mlr.press/v80/ganin18a.html) from the last year? Well, I put on my software engineering hat and now we have some code publicly available: github.com/deepmind/spiral. Go ahead and generate your own paintings!
-
Ian Goodfellow reposted
Murium
@murium9
Check out our new paper reviewing production quality ranking systems!
ceur-ws.org/Vol-2410/paper22.pdf
@KAryafar @sigir2019 @SIGIReCom #machinelearning #sigir2019 #learningtorank -
Ian Goodfellow
@goodfellow_ian
Check out the call for papers for the Bayesian deep learning workshop at NeurIPS 2019: bayesiandeeplearning.org
-
Ian Goodfellow
@goodfellow_ian
This looks like a great way to programmatically create docs with graphs etc.
Danijar Hafner
@danijarh
What I value most about Jupyter notebooks is having all results and figures together in a doc.
Today I'm releasing Python Handout, a package that lets you create docs with inline figures, images, videos directly from Python scripts. ✨ 📰 ✨
github.com/danijar/handout
Thread 👇 -
Ian Goodfellow reposted
ACM SIGGRAPH
@siggraph
Congratulations to the “GauGAN: Semantic Image Synthesis With Spatially Adaptive Normalization” team, winner of the #siggraph2019 Real-Time Live! Best in Show and Audience Choice awards! 👏 (cc: @nvidia, @cal, @MIT) #realtimelive
-
Ian Goodfellow reposted
Shakir Mohamed
@shakir_za
Really excited to share our latest paper in @nature today on machine learning for health data to make early predictions of acute kidney injury. Has been an amazing journey over the last 2 years and with an amazing set of people. www.nature.com/articles/s41586-019-1390-1
-
Ian Goodfellow reposted
shreya rajpal
@ShreyaR
I missed this when it came out, but I just discovered arxiv.org/pdf/1603.07285.pdf
The chapter on transposed convolutions is probably the best explanation I’ve seen on the subject! -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
So happy to share the news that @NVIDIADesign #gaugan won both best real time live demo award and people’s choice of best demo award in #siggraph2019 Real-Time Live. @junyanz89 @gavriilklimov @tcwang0509 @chrisjhebert Special thank to @goodfellow_ian to bring #gan to this world.
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
I will be giving a talk on #gaugan and its backbone algorithm #spade in room 501 at #siggraph2019 at 2pm today (Sunday 7/28). If you are interested in how we design a GAN to achieve the task and the idea behind it. Please come.
-
Ian Goodfellow reposted
Gracie Arenas Strittmatter
@maze518
Draw object shapes & the neural networks fill in the details like magic! @NVIDIA’s GauGAN: Semantic Image Synthesis With Spatially Adaptive Normalization - live at #siggraph2019 #realtimelive! For more: s2019.siggraph.org/conference/programs-events/real-time-l… @siggraph @chrisjhebert @junyanz89 #thriveonlive
-
Ian Goodfellow reposted
-
Ian Goodfellow
@goodfellow_ian
Keeping up my trend this week of finding out well known illustrative stories are most likely false, I found @gwern ‘s debunking of the AI rank story www.gwern.net/Tanks
-
Ian Goodfellow reposted
Daniel Ambrosi | dreamscapes.ai
@danoramas
Elements: A little "back-to-the-basics" collection inspired by one of my all-time favorite artists. Happy 149th Birthday, Maxfield Parrish! www.danielambrosi.com/LovelyLandscapes/Elements/
-
Ian Goodfellow
@goodfellow_ian
Question for psychology experts of twitter: do the “bright” rats *actually perform* better because the PhD students treat them better? Or do the PhD students report higher scores to fit their advisor’s expectations? Or both?
Ian Goodfellow
@goodfellow_ian
There’s a famous result in psychology where grad students report that rats are better at mazes if the advisor tells them the rats are “maze bright” rather than “maze dull”.
-
Ian Goodfellow
@goodfellow_ian
There’s a famous result in psychology where grad students report that rats are better at mazes if the advisor tells them the rats are “maze bright” rather than “maze dull”.
-
Ian Goodfellow
@goodfellow_ian
For years I’ve believed in the story that after the first person ran a 4 minute mile, suddenly many other athletes found that they could too, and that the 4 minute barrier was largely psychological. Today I found this article arguing that’s a myth: www.scienceofrunning.com/2017/05/the-roger-bannister-effe…
Ian Goodfellow
@goodfellow_ian
I think that barriers to the success of deep learning before 2012 were largely psychological and have previously used the 4 minute mile story as an analogy to explain it. So this surprises me.
-
Ian Goodfellow
@goodfellow_ian
DVD-GAN
AK
@_akhaliq
Efficient Video Generation on Complex Datasets
pdf: arxiv.org/pdf/1907.06571.pdf
abs: arxiv.org/abs/1907.06571 -
Ian Goodfellow
@goodfellow_ian
I’m in Fortune’s 40 under 40: fortune.com/40-under-40/2019/ian-goodfellow/
-
Ian Goodfellow reposted
Yao Qin
@YaoQin_UCSB
Very happy to release our work on detecting and diagnosing adversarial images~
Nick Frosst
@nickfrosst
@YaoQinUCSD , @colinraffel ,@sabour_sara , Gary Cottrell, @geoffreyhinton and I have released a full version of our workshop paper on capsule networks and adversarial attack detection! Check it out, or read this thread if you are busy :) 1/7
arxiv.org/pdf/1907.02957.pdf -
Ian Goodfellow
@goodfellow_ian
It’s interesting to see the pendulum swing back to representation learning. During my PhD, most of my collaborators and I were primarily interested in representation learning as a biproduct of sample generation, not sample generation itself.
Ben Poole
@poolio
BigBiGAN shows that "progress in image generation quality translates to substantially improved representation learning performance." Competitive w/self-supervised approaches on ImageNet.
The cycle from generative models to other methods and back again continues. -
Ian Goodfellow
@goodfellow_ian
While we were writing the original GAN paper, my co-author @dwf tried to get something similar to BiGAN working for representation learning. It’s pretty cool to see that work out 5 years later.
Ben Poole
@poolio
BigBiGAN shows that "progress in image generation quality translates to substantially improved representation learning performance." Competitive w/self-supervised approaches on ImageNet.
The cycle from generative models to other methods and back again continues. -
Ian Goodfellow reposted
Nick Steinmetz
@SteinmetzNeuro
As scientific teams grow, our model of credit assignment (1st author, last, or everyone else) becomes increasingly outdated. One impediment is ineffectiveness of author contributions text. Here’s a suggestion for a better way: the contributions table. A thread; feedback welcome.
-
Ian Goodfellow
@goodfellow_ian
“Be careful what you wish for”
MIT CSAIL
@MIT_CSAIL
Describe programming in only six words.
We’ll RT all the best ones.
Ours:
Turning ideas and caffeine into code.
#programmingin6words #wednesdaywisdom -
Ian Goodfellow reposted
Josh Gordon
@random_forests
A new CycleGan tutorial is ready in @TensorFlow 2.0. A very cool paper, showing how to create beautiful image transformations like:
- Horses -> zebras
- Winter -> summer
- Apples -> oranges
www.tensorflow.org/beta/tutorials/generative/cyclegan -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
In CVPR2019, the #stylegan paper won the best paper honorable mention and #spade/#gaugan paper won the best paper finalist. Congratulates to all the GAN authors! @nvidia
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
The #gaugan beta version is now available to everyone as a web service via nvda.ws/2WsY2cM #nvidia AI Playground
A short illustration video is available in www.youtube.com/watch?v=lGlJcc9HOzk
May everybody have fun with the app.
#gan, #spade -
Ian Goodfellow reposted
Peter Baylies
@pbaylies
Watch the Mona Lisa turn her head, in today's edition of stupid StyleGAN encoder tricks... #stylegan github.com/pbaylies/stylegan-encoder cc: @aarlo @Zergfriend @neurokinetikz @calamardh @roadrunning01 @Miles_Brundage
-
Ian Goodfellow
@goodfellow_ian
Some updates on this:
Ian Goodfellow
@goodfellow_ian
Whoa! It turns out that famous examples of NLP systems succeeding and failing were very misleading. “Man is to king as woman is to queen” only works if the model is hardcoded not to be able to say “king” for the last word.
Rik van Noord
@rikvannoord
1/7 Do word embeddings really say that man is to doctor as woman is to nurse? Apparently not. Check out this thread for a description of a short paper I co-wrote with Malvina Nissim and Rob van der Goot, available here: arxiv.org/abs/1905.09866 #nlproc #bias
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
2) The paper I linked to, Nissim et al 2019, also says that Bolukbasi et al 2019 solve analogies using the parallelogram math like in Mikolov et al 2013a. That's not correct. Bolukbasi et al used a different algorithm that can actually choose "queen" without hardcoding out "king"
Ian Goodfellow
@goodfellow_ian
3) Other papers have previously pointed out some of the same problems that Nissim et al 2019 point out.
-
Ian Goodfellow
@goodfellow_ian
As far as I know these kinds of analogies actually work in image models.
Ian Goodfellow
@goodfellow_ian
Whoa! It turns out that famous examples of NLP systems succeeding and failing were very misleading. “Man is to king as woman is to queen” only works if the model is hardcoded not to be able to say “king” for the last word.
Rik van Noord
@rikvannoord
1/7 Do word embeddings really say that man is to doctor as woman is to nurse? Apparently not. Check out this thread for a description of a short paper I co-wrote with Malvina Nissim and Rob van der Goot, available here: arxiv.org/abs/1905.09866 #nlproc #bias
-
Ian Goodfellow
@goodfellow_ian
Whoa! It turns out that famous examples of NLP systems succeeding and failing were very misleading. “Man is to king as woman is to queen” only works if the model is hardcoded not to be able to say “king” for the last word.
Rik van Noord
@rikvannoord
1/7 Do word embeddings really say that man is to doctor as woman is to nurse? Apparently not. Check out this thread for a description of a short paper I co-wrote with Malvina Nissim and Rob van der Goot, available here: arxiv.org/abs/1905.09866 #nlproc #bias
-
Ian Goodfellow reposted
ashwini asokan
@LadyAshBorg
#ageofgans Here’s a case of tech that can fix representation & diversity. Research shows 200-300% increase in intent to buy among black, > size 6 & many other demographics when model represents them than a std runway model.
SAY NO TO DEEPFAKES
SAY YES TO GANs done right
👩🏽👱🏽♀️👩🏽🦱👵🏼
ashwini asokan
@LadyAshBorg
Everything about GANs seems controversial these days. Here’s our CPO, Julia Dietmar @JulesKaplan in @Forbes discussing how GANs and deepfakes have the potential to create immense value for the fashion industry, when you take a new tech & put people 1st! www.forbes.com/sites/forbestechcouncil/2019/05/21/gans-an…
-
Ian Goodfellow reposted
Gamaleldin Elsayed
@gamaleldinfe
Looking forward to present my work with @jaschasd @goodfellow_ian @thisismyhat @sh_reya @NicolasPapernot @alexey2004 about adversarial examples that influence human visual perception at #vss2019 on Monday 20th (poster 43.365 Banyan Breezeway hall).
-
Ian Goodfellow reposted
David Berthelot
@D_Berthelot_ML
Another nice blog post, this time from @sanjeev_vadiraj, that describes MixMatch in a very accessible way.
medium.com/@sanjeev.vadiraj/eureka-mixmatch-a-holistic-ap… -
Ian Goodfellow reposted
Jack Clark
@jackclarkSF
Gonna be real funny in a few years when Canadian government laments the corporate capture of AI and acts like it was inevitable all along, and not easily traced back to reckless funding decisions like this. Meanwhile, China is dumping money into government-led AI initiatives.
Kate Allen
@katecallen
Recap:
1. The Ford govt cut funding for two AI research hubs
2. Its own prosperity think tank said last year it is imperative the govt keep investing in AI research to remain competitive
3. The Ford govt also cut funding for that think tank; it closed
www.thestar.com/news/gta/2019/05/19/ford-government-slash… -
Ian Goodfellow reposted
David Berthelot
@D_Berthelot_ML
Interesting blog post about semi-supervised learning by Vincent Vanhoucke
towardsdatascience.com/the-quiet-semi-supervised-revoluti… -
Ian Goodfellow reposted
Avital Oliver
@avitaloliver
Here's the code behind MixMatch, @D_Berthelot_ML's recent semi-supervised learning method.
Big improvements to common benchmarks. Basic idea:
1/ target labels for unlabeled data: avg 2 predictions of model (resample noise)
2/ sharpen target labels
3/ mixup labeled+unlabeled
David Berthelot
@D_Berthelot_ML
MixMatch (arxiv.org/abs/1905.02249) code is released github.com/google-research/mixmatch (Python3, TensorFlow 1.1x). Let me know how it works for you and also let me know if you port it to other frameworks.
-
Ian Goodfellow reposted
David Berthelot
@D_Berthelot_ML
MixMatch (arxiv.org/abs/1905.02249) code is released github.com/google-research/mixmatch (Python3, TensorFlow 1.1x). Let me know how it works for you and also let me know if you port it to other frameworks.
-
Ian Goodfellow reposted
Sergey Levine
@svlevine
Congratulations to @chelseabfinn for winning the 2019 ACM Dissertation Award!
www.acm.org/articles/bulletins/2019/may/doctoral-disserta… -
Ian Goodfellow
@goodfellow_ian
I can’t find a paper I think I remember seeing. It argued that when there is enough Lipschitz regularization all GAN losses become equivalent to WGAN, and showed that this worked for an especially silly choice of loss. Anyone know what I’m thinking of?
-
Ian Goodfellow
@goodfellow_ian
ML Twitter, what are your favorite papers / other resources about the class imbalance problem?
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Check out our new #gan work on translating images to unseen domains in the test time with few example images.
Live demo nvlabs.github.io/FUNIT/petswap.html
Project page nvlabs.github.io/FUNIT/
Paper arxiv.org/pdf/1905.01723
Video www.youtube.com/watch?v=kgPAqsC8PLM
#nvidia -
Ian Goodfellow reposted
Vaishnavh Nagarajan
@_vaishnavh
Visit poster #15 at #iclr2019 today to learn about our work (with @zicokolter) on deterministic PAC-Bayesian bounds for deep networks (and also to chat more broadly about generalization in deep learning).
-
Ian Goodfellow reposted
David Berthelot
@D_Berthelot_ML
New paper: MixMatch: A Holistic Approach to Semi-Supervised Learning arxiv.org/abs/1905.02249
Reduces error rate by up to 4x on CIFAR10.
With Nicholas Carlini @goodfellow_ian @NicolasPapernot @avitaloliver @colinraffel -
Ian Goodfellow reposted
-
Ian Goodfellow
@goodfellow_ian
If you’re at ICLR, my former colleagues will present our recent work in great hall BC today, 4:30-6:30 PM. @smnh_azadi at poster 37 (discriminator rejection sampling), @craffel at poster 46 (ACAI), and @gamaleldinfe at poster 63 (adv reprogramming)
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
@colinraffel ‘s posters: x.com/colinraffel/status/1125852224767369216
Ian Goodfellow
@goodfellow_ian
@gamaleldinfe ‘s tweet:
Gamaleldin Elsayed
@gamaleldinfe
Looking forward to present my work with @goodfellow_ian and @jaschasd on "adversarial reprogramming of neural networks" @ICLR2019 today (Poster @ Great Hall BC #63).
openreview.net/pdf?id=Syx_Ss05tm -
Ian Goodfellow reposted
Samaneh Azadi
@smnh_azadi
Looking forward to #iclr2019.
I'll present our "Discriminator Rejection Sampling" poster (#46) on Tuesday at 4:30-6:30pm in Great Hall BC.Come to our poster and get a "Cabbage Head of GAN" sticker :) @goodfellow_ian @gstsdn @catherineols @trevor_darrell_
arxiv.org/abs/1810.06758 -
Ian Goodfellow reposted
Andy Brock
@ajmooch
I'll be presenting BigGAN at ICLR2019 Monday morning, at 10:15 in Great Hall AD. I'll also be giving out Dogball stickers all throughout the conference--you can get a special shiny one if you greet me with the secret passphrase "Who's a good boy?"
-
Ian Goodfellow reposted
Victoria Krakovna
@vkrakovna
Looking forward to #iclr2019 SafeML workshop on May 6, with an awesome lineup of invited speakers and panelists @goodfellow_ian @CynthiaRudin @dhadfieldmenell @hima_lakkaraju @catherineols @rohinmshah. If you're at ICLR and interested in ML safety, join us in room R6 at 9:50am!
-
Ian Goodfellow
@goodfellow_ian
Scientific American article on GANs: www.scientificamerican.com/article/machine-learning-gets-…
-
Ian Goodfellow reposted
Mario Lucic
@MarioLucic_
Our work on "High-Fidelity Image Generation With Fewer Labels" has been accepted to ICML'19! Thanks to the reviewers and the area chairs for the thorough reviews! @icmlconf @mtschannen @TheMarvinRitter @XiaohuaZhai @OlivierBachem @sylvain_gelly @GoogleAI
-
Ian Goodfellow
@goodfellow_ian
GANs for architecture: archinect.com/news/article/150133371/would-artificial-int…
-
Ian Goodfellow
@goodfellow_ian
Robbie is a real pioneer in AI + art. Do any good art schools have a spot for this fall?
Robbie Barrat
@videodrome
Desperately trying to avoid entering another involuntary gap year... Does anyone know good art schools still accepting applications for the fall?
-
Ian Goodfellow
@goodfellow_ian
vue.ai, a startup that uses GANs to generate retail marketing images, has raised $17M: venturebeat.com/2019/04/24/vue-ai-raises-17-million-for-a…
-
Ian Goodfellow reposted
Phillip Isola
@phillip_isola
If you are interested in GANs, and around Boston on 5/31, please come check out this workshop we are organizing: ganocracy.csail.mit.edu. Talks on theory and practice, arts and applications. Also accepting poster submissions here: docs.google.com/forms/d/e/1FAIpQLSeBX-nsDXssZjHqJ2jnbM_pK…
MIT Siegel Family Quest for Intelligence (SQI)
@MIT_Quest
SAVE THE DATE! We’re co-hosting a GANs workshop @MIT on Friday 5/31 with @MITIBMLab. Tutorials, talks and posters on one of the hottest topics in AI.
ganocracy.csail.mit.edu -
Ian Goodfellow
@goodfellow_ian
Nice radio interview with @DrBeef_ : wamu.org/story/19/04/22/these-creatives-work-together-wit…
-
Ian Goodfellow reposted
CIFAR
@CIFAR_News
For #worldbookday, a selection of our favourite reads by our fellows and friends from across the physical and social sciences:
-
Ian Goodfellow
@goodfellow_ian
"Imperceptible, Robust, and Targeted Adversarial Examples for Automatic Speech Recognition" has been accepted to ICML 2019! Congratulations to @YaoQinUCSD, Nicholas, Gary, and @colinraffel ! arxiv.org/abs/1903.10346
-
Ian Goodfellow
@goodfellow_ian
TensorFuzz has been accepted to ICML 2019! Congratulations to @gstsdn !
Miles Brundage
@Miles_Brundage
"TensorFuzz: Debugging Neural Networks with Coverage-Guided Fuzzing," @gstsdn and @goodfellow_ian: arxiv.org/abs/1807.10875
-
Ian Goodfellow
@goodfellow_ian
We had a long delay between putting SAGAN on arxiv and submitting it to a conference, but it has now been accepted to ICML. Congratulations to @Han_Zhang_ @gstsdn and Dimitris!
-
Ian Goodfellow reposted
Zachary Lipton
@zacharylipton
Importance weighting is widely used but ***may have no effect on deep nets*** modulo choices regarding early stopping and weight decay. This "science-y" paper with my student Jonathon Byrd identifying this phenomenon was just accepted at #icml2019 arxiv.org/abs/1812.03372 (1/2)
-
Ian Goodfellow
@goodfellow_ian
If you’re upset that someone didn’t cite your paper, I strongly recommend contacting the authors privately before making a public complaint. Thread:
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
I’ve tried both strategies several times and can confidently say starting with private discussion is not only friendlier but also more effective.
Ian Goodfellow
@goodfellow_ian
Finally, to help change the norms of the field, I encourage you not to add citations to people who publicly attack before discussing with you.
-
Ian Goodfellow
@goodfellow_ian
OctConv is a simple replacement for the traditional convolution operation that gets better accuracy with fewer FLOPs export.arxiv.org/abs/1904.05049
-
Ian Goodfellow
@goodfellow_ian
CAIS++ (a USC student group) is holding an event to showcase their projects April 22. I had a great time visiting CAIS++ during NeurIPS 2017. If you're in the area you should check out their showcase: www.facebook.com/events/2682822088455633/
-
Ian Goodfellow reposted
Ariel Herbert-Voss
@adversariel
Watching an example of @zephoria’s “data void” search engine vulnerability materialize in real time is surreal. It bears repeating that machine learning systems are vulnerable to adversarial manipulation and the implications reach beyond media recommendation
Glen Maddern
@glenmaddern
Searched for 'Katie Bouman' cause I wanted to watch her TED talk, the fourth suggested video was this trash.
It has 1k views. On a channel with 2k subs. Why on earth is Google suggesting this to me? -
Ian Goodfellow reposted
Geoffrey Hinton
@geoffreyhinton
Thanks to my graduate students and postdocs whose work won a Turing award. Thanks to my visionary mentors Inman Harvey, David Rumelhart and Terry Sejnowski. And thanks to Jeff Dean for creating the brain team that turns basic research in neural nets into game-changing products.
-
Ian Goodfellow reposted
Andy Brock
@ajmooch
Want to train your own BigGAN on just 4-8 GPUs? Today we're proud to release BigGAN-PyTorch, a full @PyTorch reimplementation that uses gradient accumulation to get the benefits of big batches even on small hardware.
github.com/ajbrock/BigGAN-PyTorch
Repo joint work with @_alexandonian -
Ian Goodfellow reposted
Karol Kurach
@karol_kurach
Great work from Brain ZRH team on putting label-hungry GANs on a diet :)
Want to train similar GANs easily? Check out the new release of our compare_gan library! github.com/google/compare_gan
Google AI
@GoogleAI
Check out some research exploring a new approach to training conditional generative adversarial networks (GANs) that reduces the amount of labeled data required by a factor of ~10 (along with an update to the Compare GAN library!). Learn more at ↓ blog.research.google/2019/03/reducing-need-for-labeled-da…
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Check out our #cvpr19 oral paper on a new conditional normalization layer for semantic image synthesis #spade and its demo app #gaugan
paper arxiv.org/abs/1903.07291
website nvlabs.github.io/SPADE/
video1 www.youtube.com/watch?v=p5U4NgVGAwg
@tcwang0509 @junyanz89 -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Check out our #cvpr19 oral paper on a new conditional normalization layer for semantic image synthesis #spade and its demo app #gaugan
paper arxiv.org/abs/1903.07291
website nvlabs.github.io/SPADE/
video1 www.youtube.com/watch?v=p5U4NgVGAwg
@tcwang0509 @junyanz89 -
Ian Goodfellow reposted
Alexia Jolicoeur-Martineau
@jm_alexia
There's a whole community on enhancing the resolution of old #videogames mostly using ESRGAN (which uses Relativistic GANs)! 😻www.reddit.com/r/GameUpscale/
Feels good to know I am having a impact on "real" life 😺😎. -
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
Glad to see that our #gan research works enable people to "generate realistic dance videos of NBA players for in-game entertainment."
#pix2pixhd, #vid2vid
medium.com/@getxpire/how-we-used-ai-to-make-nba-players-d… -
Ian Goodfellow reposted
El Mahdi El Mhamdi
@L_badikho
Many machine learning events happening soon in north/west Africa (I list a speaker/organizer for those who want to ask questions):
Indabax:
- Cameroon (@ftasse) indabaxcameroon.github.io
- Ghana (@Moustapha_6C) www.indabaxghana.com
- Morocco (@redouanelg) indabaxmorocco.github.io/index.html -
Ian Goodfellow
@goodfellow_ian
Submissions to www.ai-safety.org are due April 12
-
Ian Goodfellow reposted
Sander Dieleman
@sedielem
Likelihood is a great loss fn, it's all about the space you measure it in! Our latest work on hierarchical AR image models (w/ @JeffreyDeFauw, Karen Simonyan): arxiv.org/abs/1903.04933
We generated 128x128 & 256x256 samples for all ImageNet classes: drive.google.com/drive/folders/1BGoEWR-tSNb4g8PkoZSp_Qh7c… (1/2) -
Ian Goodfellow
@goodfellow_ian
ACAI for modeling drum patterns: towardsdatascience.com/drum-patterns-from-latent-space-23…
-
Ian Goodfellow
@goodfellow_ian
I originally thought of GANs as an unsupervised learning algorithm, but so far, to create recognizable object categories, they've needed a supervision signal / labeled images. This new work shows how to get them to work well with few labels.
Mario Lucic
@MarioLucic_
How to train SOTA high-fidelity conditional GANs usin 10x fewer labels? Using self-supervision and semi-supervision! Check out our latest work at arxiv.org/abs/1903.02271 @GoogleAI @ETHZurich @TheMarvinRitter @mtschannen @XiaohuaZhai @OlivierBachem @sylvain_gelly
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
We are excited to share the story behind TF-Replicator, a software library that helps developers deploy their TensorFlow code between CPUs, GPUs and TPUs with minimal effort.
Blog post: deepmind.com/blog/tf-replicator-distributed-machine-learn…
Paper: arxiv.org/abs/1902.00465 -
Ian Goodfellow reposted
Dumitru Erhan
@dumierhan
VideoFlow: A Flow-Based Generative Model for Video extends Glow to a new algorithm for multi-frame video prediction with normalizing flows. Tractable & scalable! Work w/ Manoj Kumar, @babaeizadeh, @chelseabfinn, @svlevine, @laurent_dinh and @dpkingma.
arxiv.org/abs/1903.01434 -
Ian Goodfellow reposted
Jesse Engel
@jesseengel
Make music with GANs!
GANSynth is a new method for fast generation of high-fidelity audio.
🎵 Examples: storage.googleapis.com/magentadata/papers/gansynth/index.…
⏯ Colab: colab.research.google.com/notebooks/magenta/gansynth/gans…
📝 Paper: openreview.net/challenge?redirect=%2Fforum%3Fid%3DH1xQVn09FX
💻 Code: github.com/magenta/magenta/tree/main/magenta/models/gansynth
⌨️ Blog: magenta.tensorflow.org/gansynth
1/ -
Ian Goodfellow reposted
Ilya Sutskever
@ilyasut
Bill Gates highlights Dactyl as a top ten breakthrough of 2019! Congrats @woj_zaremba!! www.technologyreview.com/lists/technologies/2019/?linkId=…
-
Ian Goodfellow reposted
Gamaleldin Elsayed
@gamaleldinfe
Heading to #cosyne19 to present our work on human adversarial examples. Looking forward to meet my old friends :)
-
Ian Goodfellow
@goodfellow_ian
The arxiv paper for Lingvo is out now: arxiv.org/abs/1902.08295
-
Ian Goodfellow reposted
Negar Rostamzadeh
@negar_rz
You can apply to be an organizer of the #wiml2019 I personally learnt a lot by co-organizing the @WiMLworkshop and @WiCVworkshop 2017! It is an amazing experience and in my opinion these workshops, @WiDS_Conference @WiNLPWorkshop can eventually lead to a more inclusive community!
WiML
@WiMLworkshop
Want to help shape #wiml2019? Apply to be a #wiml2019 organizer! docs.google.com/forms/d/e/1FAIpQLScjeKOaGoT0J34tLJF3z8yXv…
Applications are due on Friday March 8th, 2019.
#machinelearning #womeninstem -
Ian Goodfellow reposted
Andy Brock
@ajmooch
BigGAN has been accepted for oral presentation at ICLR2019! We've uploaded a revision of the paper with an improved architecture, BigGAN-deep: 4x the depth, 50% *fewer* parameters, and even better performance.
openreview.net/pdf?id=B1xsqj09Fm -
Ian Goodfellow
@goodfellow_ian
More press for @lucidrains 's StyleGAN sites: www.inverse.com/article/53442-this-person-does-not-exist-…
-
Ian Goodfellow
@goodfellow_ian
This CVPR workshop is looking for applications of computer vision in developing countries. You don't need to submit a finished method paper, the workshop is actually seeking ideas for research directions and important problems to solve.
-
Ian Goodfellow
@goodfellow_ian
Philip Wang at Uber set up www.thispersondoesnotexist.com to show a new imaGANary person every time you refresh the page
-
Ian Goodfellow reposted
Victoria Krakovna
@vkrakovna
Update: the SafeML workshop paper submission deadline has been extended to March 1, for consistency with other ICLR workshops.
Victoria Krakovna
@vkrakovna
Paper submissions on a variety of AI safety topics to the #iclr2019 SafeML workshop are due Feb 22! We especially encourage submissions from underrepresented groups @WiMLworkshop @black_in_ai @_lxai #queerinai etc. CFP: sites.google.com/corp/view/safeml-iclr2019/cfp.
-
Ian Goodfellow
@goodfellow_ian
I emphatically agree. My collaborators' visa restrictions have been one of the largest bottlenecks to our collective research productivity over the last few years.
-
Ian Goodfellow reposted
Shakir Mohamed
@shakir_za
Proud that today we announce the 2019 #indabax meetings. Between March and May, Machine Learning will grow across Africa in 26 countries (from 13 last year)! Please support the work of @DeepIndaba if you can. African ML is strong and thriving 🌍 #webuildtogether #masakhane
Deep Learning Indaba
@DeepIndaba
We are proud to announce the 2019 #indabax events will take place in 26 countries across our continent! 🌍 They build leadership and strength in Machine Learning across our African continent. Please support their efforts 🎈 #webuildtogether www.deeplearningindaba.com/blog/indabax-2019-growing-the-…
-
Ian Goodfellow reposted
Yingtao Tian
@alanyttian
This is *amazingly* great in terms of quality!
-
Ian Goodfellow reposted
Mikael H Christensen
@SyntopiaDK
GAN's may be evaluated based on how smooth (disentangled) the latent space interpolations are. It is impressive how #stylegan can interpolate between different orientations - even with no concept of 3D.
-
Ian Goodfellow
@goodfellow_ian
FAIR has released code for the robust ImageNet model by Cihang Xie et al: github.com/facebookresearch/ImageNet-Adversarial-Training
-
Ian Goodfellow
@goodfellow_ian
The Toronto Deep Learning Series hosted a discussion about ACAI: www.youtube.com/watch?v=Tu3FqCD7-BY&t=718
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
We will present several #gan works in NVIDIA’s #gtc19 conference, including #stylegan, #vid2vid, and several other new GAN works that we have NOT announced. Register by 2/8 for early bird pricing and use discount code NVMINGYUL for an additional 25% off: www.nvidia.com/en-us/gtc/pricing/
-
Ian Goodfellow reposted
ICLR
@iclr_conf
Excited to preview our list of invited speakers for #iclr2019 in New Orleans:
Léon Bottou, Cynthia Dwork, Ian Goodfellow, Noah Goodman, Mirella Lapata,
Pierre-Yves Oudeyer, Emily Shuckburgh, and, Zeynep Tufekci
Tickets go on sale tomorrow @ 4pm PST at iclr.cc -
Ian Goodfellow reposted
Pushmeet Kohli
@pushmeet
Our new results on training verifiably robust neural networks for image classification.
Google DeepMind
@GoogleDeepMind
Neural networks can be vulnerable to adversarial noise. We demonstrate the remarkable effectiveness of interval bound propagation in training provably robust image classifiers: arxiv.org/abs/1810.12715
-
Ian Goodfellow reposted
Michela Paganini
@WonderMicky
Machine Learning Crash Course: Genova 🇮🇹 June 17-21. Brought to you by @MIT @UniGenova @IITalk @iaml_it
Apply: lcsl.mit.edu/courses/mlcc/mlcc2019/ #ml #ai #summerschool -
Ian Goodfellow
@goodfellow_ian
More ESRGAN mods of old games: www.pcgamer.com/modders-are-using-ai-to-upscale-pre-rende…
-
Ian Goodfellow reposted
Oriol Vinyals
@OriolVinyalsML
Happy that we could share #alphastar progress with you all! Good Games @LiquidTLO and @Liquid_MaNa, and @Artosis and @RotterdaM08 for a great show! You can see all the details in the blog.
deepmind.com/blog/alphastar-mastering-real-time-strategy-… -
Ian Goodfellow
@goodfellow_ian
I’m one of @ForeignPolicy ‘s 100 global thinkers for 2019: foreignpolicy.com/2019-global-thinkers/
-
Ian Goodfellow
@goodfellow_ian
Another great property of style-based generators is that they've finally succeeded in learning multiple levels of abstraction. Different parts of latent vector control high-level meaning vs fine-grained detail youtu.be/kSLJriaOumA?t=131
Ian Goodfellow
@goodfellow_ian
Many people including myself have tried to do this by adding noise at many different layers of the generator, but it didn't work until now
-
Ian Goodfellow
@goodfellow_ian
An exciting property of style-based generators is that they have learned to do 3D viewpoint rotations around objects like cars. These kinds of meaningful latent interpolations show that the model has learned about the structure of the world. youtu.be/kSLJriaOumA?t=335
-
Ian Goodfellow
@goodfellow_ian
Our work on adversarial examples for the human vision system has been accepted to #cosyne19 : arxiv.org/abs/1802.08195
-
Ian Goodfellow
@goodfellow_ian
4.5 years of GAN progress on face generation. arxiv.org/abs/1406.2661 arxiv.org/abs/1511.06434 arxiv.org/abs/1606.07536 arxiv.org/abs/1710.10196 arxiv.org/abs/1812.04948
-
Ian Goodfellow
@goodfellow_ian
Interpretation of a machine learning model by a human involves both the model and the human. Human misconceptions can cause as much trouble as any property of the model.
Pedro Domingos
@pmddomingos
Remarkable finding: people don't trust transparent models any more than opaque ones, and have more difficulty detecting large errors in transparent ones: arxiv.org/pdf/1802.07810.pdf
-
Ian Goodfellow reposted
👩💻 Paige Bailey
@DynamicWebPaige
✨ Applying ESRGANs to the original Doom: priceless.
-
Ian Goodfellow reposted
Catherine Olsson
@catherineols
Everyone in ML research complains about the existing peer review tools/systems.
If you're not a researcher, but you ARE someone who cares about ML as a field being sane, has a knack for product engineering, and understands communities, you could make a HUGE impact. [...] -
Ian Goodfellow
@goodfellow_ian
Super-resolution GANs for improving the texture resolution of old games: www.gamespot.com/forums/pc-mac-linux-society-1000004/esrg…
-
Ian Goodfellow reposted
Vaishnavh Nagarajan
@_vaishnavh
Uploaded our (with @zicokolter) NeurIPS 17(!) workshop spotlight paper "Generalization in Deep Networks: The Role of Distance from Initialization". We argued why it's important to take into account the initialization to explain generalization.
arxiv.org/abs/1901.01672 -
Ian Goodfellow reposted
Jen Heemstra
@jenheemstra
Mentoring myth: as you advance in your career, you outgrow the need for mentors.
As my career advances, my responsibilities are ever expanding and the stakes keep getting higher. I need more mentoring than ever and more perspectives from an increasingly diverse set of mentors. -
Ian Goodfellow reposted
Dumitru Erhan
@dumierhan
Subsidized child-care resulted in significantly more women joining the workforce and, "generated income taxes to cover more than 100 percent of the cost", thus being a program that pays for itself. A good role model for the rest of the world.
Bloomberg CityLab
@CityLab
This is how Quebec is leading the world in keeping moms in the workforce. www.bloomberg.com/news/articles/2018-12-31/affordable-day…
-
Ian Goodfellow reposted
Robbie Barrat
@videodrome
My first art show is in Paris this February; in confrontation with French painter Ronan Barrot
BARRAT/BARROT: Infinite Skulls
It's about having a neural network in a feedback loop with a painter, influencing each other's work repeatedly- and the infinitude of generative systemsThis media couldn't be displayed. See it on the original post. -
Ian Goodfellow reposted
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.....
-
Ian Goodfellow reposted
Mario Lucic
@MarioLucic_
Our work on self-modulation for GANs was accepted to #iclr2019. Essentially, it is beneficial to directly modulate batch norm as a non-linear function of the input noise. Recent works on #biggan and #stylegan inject noise in a similar fashion. arxiv.org/abs/1810.01365 @GoogleAI
-
Ian Goodfellow reposted
Gamaleldin Elsayed
@gamaleldinfe
My work with @jaschasd and @goodfellow_ian on Adversarial reprogramming of neural networks (arxiv.org/abs/1806.11146) is now accepted at #iclr2019
-
Ian Goodfellow reposted
Julian Ibarz
@julianibarz
We can now generalize to 70% grasp success on new objects with zero real data, on real robots. And we can reach 91% success by joint fine-tuning with RL with only 5k episodes. That's a 100x data efficiency gain over our QT-Opt real only baseline.
Stephen James
@stepjamUK
Sim-to-Real via Sim-to-Sim: we learn a generator that translates real-world images to a canonical simulation version to learn robot grasping with no real-world data!
arxiv.org/abs/1812.07252
w/ collaborators from @GoogleAI, @Theteamatx, @DeepMindAI -
Ian Goodfellow
@goodfellow_ian
Using GANs to learn robot grasping with no real data
Konstantinos Bousmalis
@bousmalis
RCAN: Randomized-to-Canonical Adaptation Networks are real2sim image translators trained with domain randomization requiring no real data to achieve high performance on robotic grasping, arxiv.org/abs/1812.07252 1/2
Stephen James
@stepjamUK
Sim-to-Real via Sim-to-Sim: we learn a generator that translates real-world images to a canonical simulation version to learn robot grasping with no real-world data!
arxiv.org/abs/1812.07252
w/ collaborators from @GoogleAI, @Theteamatx, @DeepMindAI -
Ian Goodfellow
@goodfellow_ian
These style-based generator results look great: arxiv.org/pdf/1812.04948.pdf
-
Ian Goodfellow
@goodfellow_ian
The Eastern European ML Community is teaching a great summer school in Romania this year. Check it out.
EEML
@EEMLcommunity
We are proud to announce the 2019 edition of EEML summer school, 1-6 July, Bucharest, Romania. Topics covered: DL, RL, computer vision, bayesian learning, medical imaging, and NLP. An amazing set of speakers confirmed so far! More info coming soon! Check www.eeml.eu!
-
Ian Goodfellow
@goodfellow_ian
Deadline extended to Dec 21
Ian Goodfellow
@goodfellow_ian
Reminder: the submission deadline for the @IEEESSP Deep Learning and Security workshop is fast approaching, Dec 14. www.ieee-security.org/TC/SPW2019/DLS/
-
Ian Goodfellow
@goodfellow_ian
#neurips2018 The number of unique z vectors seen during training turns out to be a way to control capacity and prevent memorization in GANs, separate from model size. Come see these and other insights at @_vaishnavh 's poster at 3PM today room 220D nips.cc/Conferences/2018/Schedule?showEvent=10916
-
Ian Goodfellow reposted
Karol Kurach
@karol_kurach
"Frechet Video Distance" - a new video metric well correlated with human perceived quality. Very useful if you're working on generative models for video!
We also release a set of video datasets of progressive complexity (based on Starcraft II)!
Thomas Unterthiner
@TomUnterthiner
Our paper on evaluating generative models for video is out! A cool new metric and new datasets to drive progress in this area: arxiv.org/abs/1812.01717
With @vansteenkiste_s, @karol_kurach, @RaphaelMarinier, @sylvain_gelly and other wonderful people at Google Brain Zurich. -
Ian Goodfellow reposted
Thomas Unterthiner
@TomUnterthiner
Our paper on evaluating generative models for video is out! A cool new metric and new datasets to drive progress in this area: arxiv.org/abs/1812.01717
With @vansteenkiste_s, @karol_kurach, @RaphaelMarinier, @sylvain_gelly and other wonderful people at Google Brain Zurich. -
Ian Goodfellow reposted
Michela Paganini
@WonderMicky
So excited to be at the 2nd @black_in_ai workshop at #neurips2018! 🤩
Yann Dauphin of @facebookai kicking off the event with his keynote presentation. Check out the schedule (blackinai.github.io/workshop/2018/programs/) and stop by for the poster session #neurips #blackai -
Ian Goodfellow
@goodfellow_ian
Let’s have a moment of silence at 5:10pm today, to remember the women in tech killed on this day here in Montreal 29 years ago. Thanks @AnimaAnandkumar for organizing. #neurips2018
Kristyn Wong-Tam 🇨🇦
@kristynwongtam
Today is #december6 the National Day of Remembrance and Action on Violence Against Women. On this day 29 years ago, 14 women were targeted and murdered at l'École Polytechnique de Montréal. They were separated from the men on campus and killed because they were women.
-
Ian Goodfellow
@goodfellow_ian
#neurips Check out our poster on realistic evaluation of semi-supervised learning, now until 12:45. nips.cc/Conferences/2018/Schedule?showEvent=11327
-
Ian Goodfellow reposted
DSN - Data Science Nigeria
@dsn_ai_network
#neurips2018. Inspiring one by a mentor @julius_adebayo etal on Sanity check for Saliency maps! Why? Visual inspection of explanations alone can favor methods that may provide compelling pictures,but lack sensitivity to the model &data generating process
proceedings.neurips.cc/paper_files/paper/2018/file/294a8e… -
Ian Goodfellow
@goodfellow_ian
NVIDIA gave me a new T-Rex, signed by Jensen. They are not even for sale yet! Thanks NVIDIA! The pic is with GAN extraordinaire Ming-Yu at the NVIDIA reception last night.
Ming-Yu Liu
@liu_mingyu
May the power of #titanrtx with you @goodfellow_ian
-
Ian Goodfellow
@goodfellow_ian
Come see our poster (#108) on adversarial examples that fool both computer vision and time-limited humans. Room 517a&b
-
Ian Goodfellow reposted
Feryal
@FeryalMP
It's so sad that despite all her hard work organising #wiml2018 workshop, @nunuska didn't get her visa on time to attend the workshop! I really hope for a better future...
WiML
@WiMLworkshop
Meet Nyalleng Moroosi (@nunuska ), our next organizer for #wiml2018 workshop. She is a Software Engineer and Machine Learning Researcher with the Google AI team.
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
If you want to drive a #gan in a virtual city. Please come to #nvidia booth in #neurips. We use #vid2vid to convert segmentation masks from a game engine to images in real-time. Steering wheel and pedal are provide.
youtu.be/ayPqjPekn7g -
Ian Goodfellow reposted
Demis Hassabis
@demishassabis
Proteins are essential to life. Predicting their 3D structure is a major unsolved challenge in biology and could impact disease understanding and drug discovery. I’m excited to announce that we have won the CASP13 protein folding competition! #alphafold deepmind.com/blog/alphafold/
-
Ian Goodfellow
@goodfellow_ian
Reminder: the submission deadline for the @IEEESSP Deep Learning and Security workshop is fast approaching, Dec 14. www.ieee-security.org/TC/SPW2019/DLS/
-
Ian Goodfellow reposted
WiML
@WiMLworkshop
We are excited to introduce our speakers for #wiml2018 workshop! Our first speaker is Emma Brunskill from Stanford where she leads the AI for Human Impact (@AIforHI) group! She will talk about "Reinforcement Learning to Make Good Decisions".
-
Ian Goodfellow reposted
Charles Sutton
@RandomlyWalking
Crowdsourcing research dep't: I've been asked to give a talk about how prepublication a la arxiv should affect double blind review. Does your CS conference have a prepublication policy I should know about? I know about PLDI and ACL. #neurips2018
-
Ian Goodfellow
@goodfellow_ian
I've signed this pledge and I encourage other #neurips2018 participants to do so too
Inclusion in ML
@InclusionInML
Our ML Ally pledge is up! Please consider signing:
mlally.org/pledge
#neurips2018 -
Ian Goodfellow
@goodfellow_ian
NeurIPS, Black in AI and SOCML have all lost a lot of participants due to visa issues
-
Ian Goodfellow
@goodfellow_ian
On my way to Toronto for #socml!
-
Ian Goodfellow reposted
Eric Schmidt
@ericschmidt
Climate change is real; and will be devastating to a lot of the US soon. Its much cheaper to act now than to clean up the awful mess that is going to happen. Read the report, or at least the overview. Acting now saves lives and saves money.
nca2018.globalchange.gov -
Ian Goodfellow
@goodfellow_ian
ganbreeder.app is a fun way to explore the latent space of #biggan
-
Ian Goodfellow reposted
Phillip Isola
@phillip_isola
#biggan is so much fun. I stumbled upon a (circular) direction in latent space that makes party parrots, as well as other party animals:
-
Ian Goodfellow reposted
WIRED Science
@WIREDScience
A massive new federal report describes human-caused climate change resulting in ecological collapse and infrastructural ruin by 2100. And they released it on the busiest shopping day of the year. www.wired.com/story/a-government-climate-study-contradict…
-
Ian Goodfellow reposted
Judy Gichoya
@judywawira
Registration for the 2nd @black_in_ai workshop at #neurips2018 is closing in 3 days (Deadline is 26th November 2018 at at 11:59 EST. Grab your free ticket and consider a donation to help allies and BAI members attend www.eventbrite.com/e/2nd-black-in-ai-workshop-at-nips2018…
-
Ian Goodfellow reposted
Pierre A Manzagol
@PManzagol
I wrote this document about respectful behavior at conferences. It's based on my own experience as well as discussions with others. Feedback welcome!
medium.com/@PManzagol/respectful-conferences-1b823786a7e9 -
Ian Goodfellow reposted
helena sarin
@NeuralBricolage
i'm also pleasantly surprised that they chose this piece which is actually one of my recent favorites:
GAN-powered mutants of some kind never even considered for mass production, to paraphrase Hunter Thompson -
Ian Goodfellow reposted
helena sarin
@NeuralBricolage
my artwork is featured under I - i is for… imagination 😍😊
www.bbc.com/future/gallery/20181115-a-guide-to-how-artifi… -
Ian Goodfellow reposted
nature
@Nature
Specialist visa applications in the US have fallen by one-fifth during @realDonaldTrump's presidency. go.nature.com/2QZ3UIG
-
Ian Goodfellow reposted
ACLU
@ACLU
These rules give the White House far too much discretion to avoid real scrutiny. Asking an “unauthorized” follow-up question cannot be the basis for excluding a reporter. The rules should be revised to ensure that no journalist gets kicked out of the WH for doing her job.
ErikWemple
@ErikWemple
Per White House letter to Acosta on Nov. 19, here are the new rules for press conferences at the White House.
-
Ian Goodfellow reposted
Liam Fedus
@LiamFedus
Clear @shortscienceorg paper summary by @decodyng on Language GANs Falling Short
shortscience.org/paper?bibtexKey=journals/corr/1811.02549 -
Ian Goodfellow reposted
Jeff Dean
@JeffDean
I'm very happy to see the @NipsConference board work out a great solution that keeps the full name of the conference intact but to use NeurIPS rather than NIPS (which has multiple other inappropriate connotations and was sometimes used for crude jokes at the conference). Yay!
hardmaru
@hardmaru
Announcement from @NipsConference neurips.cc/Conferences/2018/Press
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Like carbon tax, if we could somehow figure out ways to charge or credit people for their transportation choices, that would be a good thing (fewer cars on the road, more use of shared public transit, and more biking or walking with the health benefits that they provide).
Brent Toderian
@BrentToderian
What’s the full cost of this 5 km #vancouver commute?
Driving costs society $2.78.
Riding the bus costs society 38¢.
Biking saves society 75¢.
Walking saves society $1.08.
So who’s REALLY paying their way? #citymakingmath Via @thediscourse HT @grescoe www.thediscourse.ca/cities/full-cost-commute -
Ian Goodfellow reposted
Mike Tyka (@mtyka@mastodon.art)
@mtyka
Using GANs to generate Master[Finger]Prints that unlock 22-78% phones sensors (dep. on security level of sensor) arxiv.org/pdf/1705.07386.pdf .. doesn't get much more "adversarial" than that.
-
Ian Goodfellow reposted
Mihaela Rosca
@MihaelaCRosca
Wooho! Looking for a great ML summer school next year? Check out EEML, next year in Bucharest!
EEML
@TMLSS2018
Breaking news: We've decided the date and location for the next edition ... drum roll ... it will be between 1st and 6th of July 2019 in Bucharest Romania, in collaboration with @upb1818. More info coming soon.
-
Ian Goodfellow reposted
John Dearie
@John_Dearie
For years, the United States has benefitted from attracting the world's largest foreign-born student population. Now, inexplicably, the U.S. seems to have decided to intentionally forfeit that enormous strategic asset.
Bill Kerr
@william_r_kerr
"The number of new international students enrolling at American institutions fell by 6.6% during the 2017-18 academic year, on top of a 3.3% decline the year before."
www.wsj.com/articles/fewer-international-students-heading… -
Ian Goodfellow reposted
Vicki Cheung
@vmcheung
Honored to have the opportunity to share the story of #kubernetes at @lyft with the first ever #kubecon China!
It's also my first bilingual conference and the experience has been really cool! -
Ian Goodfellow reposted
Daniel Ambrosi | dreamscapes.ai
@danoramas
Swiss Dreams: A trio of scenes recently captured in Switzerland, two of which use wonderful "dreaming" styles I've never used before. Enjoy! www.danielambrosi.com/Swiss-Dreamscapes/
-
Ian Goodfellow
@goodfellow_ian
Industry titles tend to add superlatives: "research scientist" -> "senior research scientist" -> ... -> "distinguished scientist"
Ian Goodfellow
@goodfellow_ian
Academic titles tend to begin with negative adjectives and gradually remove adjectives. "Undergraduate student" -> "graduate student", "assistant professor" -> "associate professor" -> "full professor".
-
Ian Goodfellow
@goodfellow_ian
The "assistant professor" title seems especially galling to me: who exactly is the assistant professor assisting? They do the full job.
Ian Goodfellow
@goodfellow_ian
Academic titles tend to begin with negative adjectives and gradually remove adjectives. "Undergraduate student" -> "graduate student", "assistant professor" -> "associate professor" -> "full professor".
-
Ian Goodfellow
@goodfellow_ian
Academic titles tend to begin with negative adjectives and gradually remove adjectives. "Undergraduate student" -> "graduate student", "assistant professor" -> "associate professor" -> "full professor".
-
Ian Goodfellow
@goodfellow_ian
Using differentiable ray tracing to make 3D adversarial example objects: people.csail.mit.edu/tzumao/diffrt/
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
The BigGAN generators from our paper arxiv.org/abs/1809.11096 are now available on TF Hub (tfhub.dev/s?q=biggan). Try the Colab demo at: colab.research.google.com/github/tensorflow/hub/blob/mast…
-
Ian Goodfellow
@goodfellow_ian
Adversarial examples for perceptual ad-blocking: floriantramer.com/docs/papers/misc18adblock.pdf
-
Ian Goodfellow reposted
Wieland Brendel
@wielandbr
Results of the NIPS Adversarial Vision Challenge are in! Common themes are low-frequency versions of the boundary attack, ensembling of defenses and new adversarial training schemes. Find all the details at medium.com/bethgelab/results-of-the-nips-adversarial-visi…
-
Ian Goodfellow
@goodfellow_ian
An updated 2 year progress pic for ImageNet GANs. New pic by @gstsdn includes latest results by @ajmooch et al.
-
Ian Goodfellow reposted
GeekPwn
@GeekPwn
CAAD CTF was held on #geekpwn2018 Shanghai. 6 teams are invited. Finally team IYSWIM(Yuxin Wu, Cihang Xie) won 1st place prize RMB 30K. They successfully attacked online celebrity recognition service of Amazon @awscloud. Make it think the presenter is Arnold @Schwarzenegger.
-
Ian Goodfellow reposted
GeekPwn
@GeekPwn
The Competition on Adversarial Attacks and Defenses (CAAD 2018) ended. 5 top teams of 3 sub-competiitons shared total prize USD $100K. Congratulate team ysharma1126 won 1st places of attacks and team FAIR & JHU won 1st place of Defense. More at: hof.geekpwn.org/caad/en
-
Ian Goodfellow reposted
Noam Brown
@polynoamial
I'm honored to share the second-ever @IJCAIconf Marvin Minsky Medal with Professor Sandholm for our work on Libratus, the AI that beat top pros in no-limit poker! Here's a high-level explanation of the AI: www.youtube.com/watch?v=2dX0lwaQRX0
-
Ian Goodfellow reposted
Pasquale Minervini
@PMinervini
At #emnlp18 there was a lot of interest in adversarial examples - inputs designed to cause your model to make a mistake. I think that most people in NLP only see them as slightly perturbed inputs where predictions are different (as in vision), but this is just a special case. 1/3
-
Ian Goodfellow reposted
matt blaze
@mattblaze
There’s currently an organized bot effort to discourage people from voting. Among their “arguments” is that voting machines are all “rigged”.
Hogwash.
I’m one of the computer scientists who discovered these voting system problems. And I vote, because it matters. So should you. -
Ian Goodfellow reposted
Catherine Olsson
@catherineols
Ok, I realize now that
if you didn't do half a PhD studying human vision with fMRI, this quote doesn't make sense; and
"OMG" isn't an explanation;
also @goodfellow_ian is messaging me on chat asking great questions;
so let me just broadcast an explanation of why "OMG": 1/n
Catherine Olsson
@catherineols
OMG:
"We scanned with fMRI a unique group of adults who, as children, engaged in extensive experience with a novel stimulus, Pokemon. [...] the experienced retinal eccentricity during childhood predicts the locus of distributed responses to Pokemon in adulthood." -
Ian Goodfellow reposted
Sundar Pichai
@sundarpichai
We’re announcing our AI for Social Good program, applying @GoogleAI expertise to projects w/ positive societal impact. We invite researchers & orgs to submit ideas for the #googleai Impact Challenge: @googleorg will fund $25M in grants to chosen projects www.blog.google/outreach-initiatives/google-org/ai-social…
-
Ian Goodfellow reposted
Oriol Vinyals
@OriolVinyalsML
Between the no name change decision and not doing more to ensure a more fair distribution of the waitlist tickets (other than a FIFO queue), I am quite disappointed @NipsConference today! #protestnips
-
Ian Goodfellow reposted
Tejas Kulkarni
@tejasdkulkarni
Why are polls being used here to make the final decision? Isn't it common sense to change the name to create a more inclusive vibe in the community?
-
Ian Goodfellow reposted
Laurent Dinh
@laurent_dinh
On a sidenote: why are people so attached to the @NipsConference name? The lab I'm from (@MILAMontreal) changed name midway and I swear it was fiiiiine 😂
-
Ian Goodfellow reposted
Ben Hamner
@benhamner
The @NipsConference “decision by committee” to keep its name is pathetic. We do research because we want to improve the world a decade for now, not tomorrow. @NipsConference should choose its name based on where it wants the field to be in a decade as well
Jeff Dean
@JeffDean
I think enough people are made to feel uncomfortable by the current name that the NIPS board should change the name of the @NipsConference. #protestnips
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
I think enough people are made to feel uncomfortable by the current name that the NIPS board should change the name of the @NipsConference. #protestnips
-
Ian Goodfellow reposted
Barbara Engelhardt
@BeEngelhardt
ML @mxlearn comments about @NipsConference not changing its name are despicable. E.g., "We won this round, now it's up to us to make sure the newly installed diversity panels are mocked relentlessly and not given the legitimacy they crave." www.reddit.com/r/MachineLearning/comments/9qtzck/n_nips_k…
-
Ian Goodfellow
@goodfellow_ian
I think changing the name of NIPS is the right thing to do. The majority of women in the poll voted for it, and moral leadership shouldn’t be driven by polls anyway.
-
Ian Goodfellow reposted
Negar Rostamzadeh
@negar_rz
Considering the majority vote when the minorities are mainly affected doesn't sound right to me! @NipsConference
-
Ian Goodfellow
@goodfellow_ian
Interested in jump-starting your career in machine learning research? Consider the Google AI Residency Program! Applications are now open until January 28th, 2019! Check out g.co/airesidency/apply for more information.
-
Ian Goodfellow reposted
Sam Altman
@sama
We need to research more aggressive climate solutions, and YC is ready to fund: carbon.ycombinator.com
-
Ian Goodfellow
@goodfellow_ian
Six amazing guest editors are running a special issue of IJCV on GANs. Submissions due in March 2019. More info here: people.csail.mit.edu/junyanz/ijcvgans.html
-
Ian Goodfellow reposted
Inclusion in ML
@InclusionInML
We've just updated the nips2018inclusion.weebly.com/childcare.html webpage to include links for outside childcare providers (recommended by local hotels) and activities for children (thanks to colleagues in Montreal for the suggestions)!
Please let us know if you know any other good resources! -
Ian Goodfellow reposted
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.
-
Ian Goodfellow reposted
Samaneh Azadi
@smnh_azadi
Apply this simple rejection sampling technique to get rid of the bad samples generated by your GAN model.
-
Ian Goodfellow
@goodfellow_ian
Both @catherineols and I set our Google employee profile pictures to the cabbage head sample for a few days
Catherine Olsson
@catherineols
I gotta say, my absolute favorite thing from the Discriminator Rejection Sampling paper is this cabbage-head GAN sample from Figure 4 😂
arxiv.org/pdf/1810.06758.pdf -
Ian Goodfellow
@goodfellow_ian
The discriminator often knows something about the data distribution that the generator didn't manage to capture. By using rejection sampling, it's possible to knock out a lot of bad samples.
-
Ian Goodfellow
@goodfellow_ian
PSA. If you’re a US green card holder, requirements to travel to Canada have changed recently. I imagine many of my followers are headed to NIPS / SOCML soon
Kamelia Aryafar
@KAryafar
If you are a green card holder remember that you need an ETA before traveling to Canada ! I didn’t know this and just missed a flight to Canada , it’s a new law 🧐
-
Ian Goodfellow reposted
Dickey Singh
@DickeySingh
If you have not read @goodfellow_ian’s Deep learning book and would like to read it with others, see this blackswans.io/post/44/ and retweet to those who may want to join the study group.
-
Ian Goodfellow reposted
Shreya Shankar
@sh_reya
disclaimer: I’ve never been an AI resident, but I always get questions about the residency program. I’ve learned that most people think it’s for PhDs and experts, but Google actually consciously accepts people from many different levels + backgrounds
Google AI
@GoogleAI
Applications for the 2019 Google AI Residency program are now open! Visit g.co/airesidency/apply for more information on how to apply. To learn more about the accomplishments of the recently graduated second class of residents, visit ↓ research.google/blog/highlights-from-the-google-ai-reside…
-
Ian Goodfellow reposted
Hugo Larochelle
@hugo_larochelle
It's now possible to apply for the 2019 Google AI Residency Program. We'll be taking residents at many locations, including Montreal :-)
Google AI
@GoogleAI
Applications for the 2019 Google AI Residency program are now open! Visit g.co/airesidency/apply for more information on how to apply. To learn more about the accomplishments of the recently graduated second class of residents, visit ↓ research.google/blog/highlights-from-the-google-ai-reside…
-
Ian Goodfellow reposted
Hector Yee
@eigenhector
Google 2019 research internships are now open careers.google.com/jobs
-
Ian Goodfellow reposted
Andy Brock
@ajmooch
GANs learn about eyes very early in training, and immediately begin sticking them everywhere. I'd like to think this one would make a good album cover.
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
We present large-scale training of generative models, which sets the new state of the art in image synthesis on ImageNet: arxiv.org/abs/1809.11096
-
Ian Goodfellow reposted
Jun-Yan Zhu
@junyanz89
We are hosting an IJCV Special Issue on Generative Adversarial Networks for Computer Vision (@elishechtman @liu_mingyu @jankautz ). [Due: 31 Mar 2019]. Here is the Call for Papers: people.csail.mit.edu/junyanz/ijcvgans.html
-
Ian Goodfellow reposted
Sergey Levine
@svlevine
Imitation directly from images, inverse RL with with transferable rewards, and better GANs can all be enabled by adding an information bottleneck into the discriminator!
Jason Peng
@xbpeng4
New paper on regularizing adversarial learning with a variational discriminator bottleneck. Our method (VAIL) can imitate challenging skills from mocap data, and can even learn to run directly from a video clip:
xbpeng.github.io/projects/VDB/index.html -
Ian Goodfellow reposted
Andy Brock
@ajmooch
Large-Scale GAN Training: My internship project with Jeff and Karen.
We push the SOTA Inception Score from 52 -> 166+ and give GANs the ability to trade sample variety and fidelity.
arxiv.org/abs/1809.11096 -
Ian Goodfellow reposted
Jeremy Howard
@jeremyphoward
We have fun with WGAN in course.fast.ai, and I challenge students to try to get something working that's even better. Some nice results here from a student using SAGAN+D2GAN+SNGAN, showing a "woman->man" latent space :)
forums.fast.ai/t/try-different-gans-sagan-d2gan-sngan/23626 -
Ian Goodfellow
@goodfellow_ian
If you made a big neural net or dataset recently, let me know the # of examples / “neurons” / average # connections per neuron. I want to update my graphs of how these change over time
Nando de Freitas
@NandoDF
ICLR 2019 lessons thus far: The deep neural nets have to be BIGGER and they’re hungry for data, memory and compute. GANs, Res-blocks, LSTMs, convnets, & multiagent tricks are doing the job.
-
Ian Goodfellow reposted
Oriol Vinyals
@OriolVinyalsML
Best GAN samples ever yet? Very impressive ICLR submission! BigGAN improves Inception Scores by >100.
Paper: openreview.net/pdf?id=B1xsqj09Fm
Lots more samples: drive.google.com/drive/folders/1lWC6XEPD0LT5KUnPXeve_kWeY… -
Ian Goodfellow
@goodfellow_ian
The current XKCD ( xkcd.com/2050/ ) jokingly proposes a time system where the length of a second changes throughout the year. Fun fact: there really was a time system like this: en.wikipedia.org/wiki/Japanese_clock
-
Ian Goodfellow reposted
Coefficient Giving
@coeff_giving
Applications are now open for our second annual AI Fellows Program to support promising PhD students in #ai and #machinelearning! Learn more here: www.openphilanthropy.org/focus/global-catastrophic-risks/…
-
Ian Goodfellow reposted
The Gradient
@gradientpub
“As my watercolor teacher used to say: let the medium do it. My sketch simply provides the foundation, and then I let the network do its thing.” Helena Sarin, an artist who works with GANs, reflects on her creative process:
thegradient.pub/playing-a-game-of-ganstruction/ -
Ian Goodfellow reposted
Google AI
@GoogleAI
In order to incentivize and measure progress towards the goal of zero confident classification errors in #machinelearning models, we're announcing the Unrestricted Adversarial Examples Challenge. Learn how to participate in the blog post below! blog.research.google/2018/09/introducing-unrestricted-adv…
-
Ian Goodfellow reposted
Chelsea Finn
@chelseabfinn
Enjoyed giving a 3 min talk today at #emtechmit, by @techreview, and meeting [some of] the other 34 under 35!
Check out my talk here (and a bunch of other awesome talks): www.technologyreview.com/video/612036/2018-innovator-unde… -
Ian Goodfellow reposted
Vicki Cheung
@vmcheung
Excited to keynote Kubecon again!
Kubernetes has matured so much since my last Kubecon. There's going to be lots to talk about! :)
CNCF
@CloudNativeFdn
#kubecon + #cloudnativecon China News: visit www.cncf.io/blog/2018/09/11/announcing-the-final-four-key… to check out the final four Shanghai keynotes from speakers like @vmcheung, @brendandburns + more 🇨🇳
-
Ian Goodfellow reposted
Shreya Shankar
@sh_reya
I’m super stoked to launch Trace, a educational comics platform for computer science. Check it out below!
If you’re excited by our mission to make learning computer science simple and fun, please like, RT, or share your favorite comic strip. Thanks so much!
Trace Comics
@trace_comics
That feel when you’re on a totally different wavelength...
Learn about artificial intelligence with Trace at www.trace.fyi.
#ai #technology #education #comics -
Ian Goodfellow reposted
Fernanda Viégas
@viegasf
Check out PAIR's What-If Tool: Code-Free Probing of Machine Learning Models. It allows you to test for multiple ML Fairness measures:
ai.googleblog.com/2018/09/the-what-if-tool-code-free-prob… -
Ian Goodfellow reposted
Adrià Gascón
@adria
We’re organizing a privacy workshop on Dec. 8th at #nips2018 :)
Invited speakers include @goodfellow_ian @kamalikac and Shafi Goldwasser. Deadline for submissions is early October.
All details at ppml-workshop.github.io/ppml/ -
Ian Goodfellow reposted
Ariel Herbert-Voss
@adversariel
Excellent discussion about using differential privacy for actual situations beyond university whiteboards. Despite deployment challenges, diffpriv still remains an effective method for protecting sensitive data and everyone using ML should keep it in their toolbox
Miles Brundage
@Miles_Brundage
"Issues Encountered Deploying Differential Privacy," Garfinkel et al.: arxiv.org/abs/1809.02201
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
We will be presenting our work on Multimodal Unsupervised Image-to-Image Translation (MUNIT) on P-1B-06 from 4-6pm on Monday at #eccv2018 Please drop by if you want to know more details about our work. Code is available at github.com/NVlabs/MUNIT
-
Ian Goodfellow reposted
DSN - Data Science Nigeria
@dsn_ai_network
We welcome the world to Africa’s biggest #machinelearning fiesta, @DeepIndaba.
The creme de la creme of the AI space are here to share best practice. Pls follow the schedule by clicking www.deeplearningindaba.com/schedule-2018.html
Poster Schedule, click goo.gl/jxtN5a
@shakir_za @kbeguir -
Ian Goodfellow reposted
Yaroslav Bulatov
@yaroslavvb
The code to reproduce "Imagenet in 18 minutes" is posted github.com/diux-dev/imagenet18 , please let me know if you run into any problems. It actually becomes "Imagenet in 12 minutes" if we use 74.9% top1 used in Chainer's "Imagenet in 15" paper, last few bits are the hardest.
-
Ian Goodfellow reposted
Koichi Hamada
@hamadakoichi
Full-body High-resolution Anime Generation with Progressive Structure-conditional GANs.
- Full-body HD anime generation.
- Structural conditions on both the generator and discriminator at each scale.
Project Page: dena.com/intl/anime-generation Generated Anime: youtu.be/bIi5gSITK0E -
Ian Goodfellow
@goodfellow_ian
Congratulations to @julius_adebayo , @_beenkim, @muelly and @mrtz ! "Sanity Checks for Saliency Maps" (on arxiv soon) will be a spotlight at NIPS 2018
-
Ian Goodfellow
@goodfellow_ian
Congratulations to @avitaloliver , @gstsdn , @colinraffel , and @ekindogus ! "Realistic Evaluation of Deep Semi-Supervised Learning Algorithms" will be a spotlight at NIPS 2018. arxiv.org/abs/1804.09170
-
Ian Goodfellow
@goodfellow_ian
Congratulations to @gamaleldinfe , @sh_reya , @thisismyhat , @NicolasPapernot , @alexey2004 and @jaschasd ! "Adversarial Examples that Fool both Computer Vision and Time-Limited Humans" arxiv.org/abs/1802.08195 was accepted to NIPS 2018.
-
Ian Goodfellow
@goodfellow_ian
The third Self-Organizing Conference on Machine Learning will be held in Toronto, November 30-December 1. Get more information and apply to attend here: sites.google.com/corp/view/socml-2018/application
-
Ian Goodfellow reposted
Hanna Wallach (@hannawallach.bsky.social)
@hannawallach
Heads-up #nips208 attendees: there are childcare grants available this year!
Katherine Heller
@kat_heller
There are child care grants at nips this year. Please tell your friends and colleagues.
-
Ian Goodfellow
@goodfellow_ian
MIT Technology Review article on Everybody Dance Now: www.technologyreview.com/s/611941/easy-to-make-videos-can…
-
Ian Goodfellow reposted
WithTheBest
@WithTheBest
Announcing our final #aiwtb keynote, Ian Goodfellow (@goodfellow_ian), Research Scientist at Google Brain!
Use AIWTB50 for 50% off your virtual conference ticket! You'll get access to replays on each talk from our awesome #ai thought leaders goo.gl/8WnQ4y -
Ian Goodfellow
@goodfellow_ian
www.nature.com/articles/s41586-018-0438-y Deep learning for predicting aftershocks of large earthquakes. Besides offering better predictions, interpretations of the model suggest promising directions for new physical theories
-
Ian Goodfellow
@goodfellow_ian
Investigation of GAN mode collapse revealed that datasets mode collapse too
-
Ian Goodfellow
@goodfellow_ian
Thread from a biostatistician at a hospital about how hard it is to participate in the AI research community while neither in academia nor industry
Alexia Jolicoeur-Martineau
@jm_alexia
Panelists of the #montreal #ai #ecoystem @froginthevalley @marktmaclean @MILAMontreal @rllabmcgill @NarjesBoufaden @PhilBeaudoin of @SymposiumAi, we need to talk. I’m bringing other AI researchers (from Montreal or not) into the discussion because I feel it’s very important. 1/15
-
Ian Goodfellow
@goodfellow_ian
Security researchers show that they are able to remotely move a robot in a university research lab. Many robots accidentally expose publicly available connections to the internet. arxiv.org/abs/1808.03322
-
Ian Goodfellow
@goodfellow_ian
I don’t like notebooks either. I agree with a lot of @joelgrus ‘s reasons and have a few others of my own
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
A notebook means I need to set up my environment correctly in two places rather than one (front end and back end)
Ian Goodfellow
@goodfellow_ian
Notebooks often degrade worse on flaky shuttle / airplane WiFi than ssh / screen do
-
Ian Goodfellow
@goodfellow_ian
GANs for imitating dance moves
-
Ian Goodfellow
@goodfellow_ian
A Portrait of the Artist as a Young GAN
Shreya Shankar
@sh_reya
it's ok, little generative model. i know you are confused about your identity, but you are not a beach. hopefully one day you will learn
(description: using attentional GANs to generate images from text, demo: t2i.cvalenzuelab.com, paper: arxiv.org/abs/1711.10485) -
Ian Goodfellow
@goodfellow_ian
Video-to-video translation with conditional GANs
Ming-Yu Liu
@liu_mingyu
-
Ian Goodfellow
@goodfellow_ian
Blog post on adversarial reprogramming: rajatvd.github.io/Exploring-Adversarial-Reprogramming/
-
Ian Goodfellow reposted
Hao Li
@HaoLi81
Check out @PinscreenInc's SIGGRAPH 2018 Real Time Live! Presentation showcasing paGAN, our photoreal avatar machine from hell: youtu.be/rPam5CHFQMQ?t=4562
-
Ian Goodfellow
@goodfellow_ian
Evaluating generative models by estimating a latent skill variable that explains their performance seems like a promising new research direction
Catherine Olsson
@catherineols
Our paper "Skill Rating for Generative Models" is now up! arxiv.org/abs/1808.04888
tl;dr: A new idea & proof-of-concept for evaluating generative models. Train a bunch of GANs. Have the generators "play against" all the discriminator snapshots. Rate them like chess players. 1/n -
Ian Goodfellow
@goodfellow_ian
This anonymous article is part of the conversation around the NIPS name change: medium.com/@lupopolvuh/the-harm-of-harmless-jokes-3ac5c65…
-
Ian Goodfellow reposted
Nando de Freitas
@NandoDF
As someone who made a career publishing at NIPS since 97, I’m delighted that the acronym is changing. As a dad, I’m also looking forward to much needed childcare while attending the conference.
Zachary Lipton
@zacharylipton
Whether or not anyone supports the NIPS name change, there's no excuse for dismissing or making ad hominem attacks against the women asking for it. Also the other bad behavior described here is wack... a workshop that gives out "Romeo & Juliet" awards? medium.com/@lupopolvuh/the-harm-of-harmless-jokes-3ac5c65…
-
Ian Goodfellow
@goodfellow_ian
Train ImageNet in 18 minutes for just $40. www.fast.ai/2018/08/10/fastai-diu-imagenet/ By my former colleague @yaroslavvb
-
Ian Goodfellow
@goodfellow_ian
One of the most anticipated GAN names now taken: paGAN. www.fxguide.com/featured/a-i-at-siggraph-part-2-pinscreen…
Ian Goodfellow
@goodfellow_ian
Also, someone DMed me to warn me these results could be fake: www.reddit.com/r/MachineLearning/comments/8zm4kl/d_lawsui…
Ian Goodfellow
@goodfellow_ian
To be clear, we don’t know if these allegations are true, just that they exist. And the company probably can’t comment publicly while the litigation is pending
Ian Goodfellow
@goodfellow_ian
A few more people have contacted me privately to say that these allegations are false / Pinscreen's results are real. Some of the people who have contacted me privately are people I know from before from other contexts who don't have a stake in Pinscreen
-
Ian Goodfellow
@goodfellow_ian
This is how @gstsdn is able to test if some coverage is new coverage without explicitly storing all old coverage for TensorFuzz. arxiv.org/abs/1807.10875
-
Ian Goodfellow
@goodfellow_ian
Congratulations, @junyanz89 !
Berkeley AI Research
@berkeley_ai
BAIR PhD graduate @junyanz89 has received the 2018 SIGGRAPH Dissertation Award! Congratulations!
www.siggraph.org/outstanding-doctoral-dissertation-award-… -
Ian Goodfellow
@goodfellow_ian
I’m hoping this will become the basis of regression testing for complicated software involving ML. eg, before rolling out the new version, fuzz searching for discrepancies between new and old version
Ian Goodfellow
@goodfellow_ian
arxiv.org/pdf/1807.10875.pdf TensorFuzz automates the process of finding inputs that cause some specific testable behavior, like disagreement between float16 and float32 implementations of a neural network
Ian Goodfellow
@goodfellow_ian
Neural networks are notoriously hard to debug. @gstsdn has developed a new debugging methodology by adapting traditional coverage guided fuzzing techniques to neural networks.
-
Ian Goodfellow
@goodfellow_ian
arxiv.org/pdf/1807.10875.pdf TensorFuzz automates the process of finding inputs that cause some specific testable behavior, like disagreement between float16 and float32 implementations of a neural network
Ian Goodfellow
@goodfellow_ian
Neural networks are notoriously hard to debug. @gstsdn has developed a new debugging methodology by adapting traditional coverage guided fuzzing techniques to neural networks.
-
Ian Goodfellow
@goodfellow_ian
Neural networks are notoriously hard to debug. @gstsdn has developed a new debugging methodology by adapting traditional coverage guided fuzzing techniques to neural networks.
-
Ian Goodfellow
@goodfellow_ian
Causal InfoGAN imagining plans for how to move a rope into different positions: arxiv.org/abs/1807.09341
-
Ian Goodfellow
@goodfellow_ian
I suspect that peer review *actually causes* rather than mitigates many of the “troubling trends” recently identified by @zacharylipton and Jacob Steinhardt: arxiv.org/abs/1807.03341
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
(This last one is a bit less consistent than the others. I’ve seen a few science papers get high review scores without having to sell out... but they often get rejected from a few conferences before being lucky enough to get reviewers who get it)
Ian Goodfellow
@goodfellow_ian
Some of the other troubling trends would probably happen without peer review, but I see reviewers basically asking to add mathiness, spurious explanations, and spurious novelty all the time
-
Ian Goodfellow reposted
Ryan Delk
@delk
It’s hard to overstate how important Talent Magnets are for building great teams.
Hiring people who draw top talent in to work with them is a super power, especially early for startups. -
Ian Goodfellow reposted
Black in AI
@black_in_ai
We're excited to announce the Call for Participation for the 2nd Black in AI Workshop at #nips2018! Submission deadline is August 30th & travel grant apps are due August 24th. blackinai.github.io
Stay tuned for program announcement. Hope to see you all in Montreal @ #nips! -
Ian Goodfellow reposted
Gene Kogan
@genekogan
looks like i found the meme vector
-
Ian Goodfellow
@goodfellow_ian
Also www.deeplearningbook.org Thank you @matplotlib !
Kacper Kowalik
@Xarthisius
. @matplotlib this is just a teaser for things to come: You made (at least) 13% of arXiv publications possible last month and 11.3% of all arXiv publications in 2018. Congrats! #scipy2018 #sprints
-
Ian Goodfellow reposted
Mihaela Rosca
@MihaelaCRosca
Really excited to be part of #tmlss. The summer school contains a set of great speakers and tutorials and starts today in Cluj, România. More at tmlss.ro
-
Ian Goodfellow
@goodfellow_ian
Also, area chairs: if you see reviewers making criticisms that are too vague to be meaningfully addressed in the rebuttal, ask the reviewer to make the review more specific. If nothing else you probably need this level of detail to write a good meta-review.
Prof. Akiko Iwasaki
@VirusesImmunity
A referee was condescending and negative of our paper stating, “many errors, too many to detail”. After we pressed for examples, he/she was unable to point to a single one.
Reviewers, if you cannot back up your criticisms with facts, don’t make them. Authors, #callthemout -
Ian Goodfellow reposted
Alexey Kurakin
@alexey2004
Training and evaluation code for "Adversarial logit pairing" paper is released: github.com/tensorflow/models/tree/master/research/adversa…
@goodfellow_ian @harinidkannan -
Ian Goodfellow reposted
TensorFlow
@TensorFlow
This paper includes 15 TensorFlow Hub modules to generate images using GANs! Check out the colab notebook to try it out → colab.research.google.com/github/google/compare_gan/blob/… #tfhub
Google AI
@GoogleAI
Attending #icml2018? Visit the Reproducibility Workshop (mltrain.cc/events/enabling-reproducibility-in-machine-lea…) to hear new work from the Google Brain team on a practical “cookbook” for #gan research (arxiv.org/abs/1807.04720), along with AI Resident work on new educational framework, www.depthfirstlearning.com
-
Ian Goodfellow reposted
‘cisca O Oladipo, PhD, FBCS
@ciscaoladipo
Beautiful piece on our Timnit!
@timnitGebru
AI Innovators: How One Woman Followed Her Passion and Brought Diversity to AI via @forbes www.forbes.com/sites/nvidia/2018/07/10/ai-innovators-how-… -
Ian Goodfellow
@goodfellow_ian
I see your joke suggestion, and raise you "Icy ML"
Tim Vieira
@xtimv
New name for @NipsConference "AI Winter" — Miro Dudík
-
Ian Goodfellow reposted
Catherine Olsson
@catherineols
UPDATE: Times for the @iclmconf debates workshop have shifted earlier to accommodate the World Cup final! www.machinelearningdebates.com/program
11:30-12:30: Rigor
1:30-2:30: Security
2:30-3:30: Fairness
4:00-5:00: Deep Learning
5:00-6:00: ⚽️ France vs. Croatia ⚽️ -
Ian Goodfellow
@goodfellow_ian
The latest from the team that brought you "Are GANs created equal?" The new cookbook now endorses some GAN techniques as better than others.
Karol Kurach
@karol_kurach
New work from our group @GoogleAI - A practical “cookbook” for #gan research in 2018: arxiv.org/abs/1807.04720
Code & pre-trained TensorFlow Hub models: github.com/google/compare_gan
Attending #icml2018 ? Visit our poster on Saturday (Reproducibility Workshop) -
Ian Goodfellow reposted
Tejas Kulkarni
@tejasdkulkarni
If you are curious about how DeepRL, GANs and program synthesis fit together in the long term, come to our ICML oral at 5 pm in A7 (deep learning adversarial section)
Tejas Kulkarni
@tejasdkulkarni
Work with @yaroslav_ganin , Igor, @arkitus and @OriolVinyalsML. DeepRL + GANs + Program Synthesis + Robots
Oriol Vinyals
@OriolVinyalsML
RL + GANs: Program synthesis with an agent that uses a paint program to fool a discriminator. Paper+Blog: deepmind.com/blog/learning-to-generate-images/
-
Ian Goodfellow reposted
Wieland Brendel
@wielandbr
Do you have an idea on how to attack neural network models using only a small number of queries? Or do you have an idea on how to defend against adversarial attacks? Join our NIPS 2018 Adversarial Vision Challenge! www.crowdai.org/organizers/bethgelab/challenges/nips-2018….
-
Ian Goodfellow reposted
Google AI
@GoogleAI
A lot of @TensorFlow demos at the #icml2018 Google booth today!
•TF Hub (goo.gl/rt5e2N) 10:30-11:00
•TF AutoGraph (github.com/tensorflow/tensorflow/blob/master/tensorflow/c…) & TF AdaNet (ai.google/research/pubs/pub46541/) 12:00-13:30
•TF Transform (research.google/blog/preprocessing-for-machine-learning-w…) 15:30-16:00
•TF Adanet: 18:15-19:15 -
Ian Goodfellow reposted
Google AI
@GoogleAI
Long day of #icml2108 talks? Stop by the Google booth from 15:30-16:00 to have fika (en.wikipedia.org/wiki/Fika_(Sweden)) and learn more about Tacotron (blog.research.google/2018/03/expressive-speech-synthesis-…) with Software Engineers @daisystanton & @EricBattenberg. Oh and TF Transform (research.google/blog/preprocessing-for-machine-learning-w…) too!
-
Ian Goodfellow
@goodfellow_ian
I'll teach an ACM webinar on adversarial machine on July 24: event.on24.com/eventRegistration/EventLobbyServlet?target…
-
Ian Goodfellow
@goodfellow_ian
Nice GAN results at ICML: avg.is.tuebingen.mpg.de/publications/meschedericml2018
Lars Mescheder
@LMescheder
@MPI_IS_Tue - Happy to present our recent work on GAN stability at #icml2018. Nice side result: High resolution (1024x1024) GAN without progressively growing architectures or advanced normalization layers. avg.is.tuebingen.mpg.de/research_projects/convergence-and…
www.youtube.com/watch?v=GY3nPi96ZgI -
Ian Goodfellow
@goodfellow_ian
Adversarial examples for the GRE
Ben Zimmer
@bgzimmer
This gobbledygook earns a perfect grade from the GRE's automated essay scoring system. Algorithms writing for algorithms. www.npr.org/2018/06/30/624373367/more-states-opting-to-ro…
-
Ian Goodfellow
@goodfellow_ian
This new family of GAN loss functions looks promising! I'm especially excited about Fig 4-6, where we see that the new loss results in much faster learning during the first several iterations of training. I implemented the RSGAN loss on a toy problem and it worked well.
Alexia Jolicoeur-Martineau
@jm_alexia
My new paper is out! " The relativistic discriminator: a key element missing from standard GAN" explains how most GANs are missing a key ingredient which makes them so much better and much more stable! #deeplearning #ai
ajolicoeur.wordpress.com/RelativisticGAN/
arxiv.org/abs/1807.00734 -
Ian Goodfellow reposted
Jascha Sohl-Dickstein
@jaschasd
Adversarial Reprogramming of Neural Networks arxiv.org/pdf/1806.11146 A new goal for adversarial attacks! Rather than cause a specific misclassification, we force neural networks to behave as if they were trained on a completely different task! With @gamaleldinfe, @goodfellow_ian
-
Ian Goodfellow
@goodfellow_ian
Congratulations Chelsea! Tremendously deserved
Chelsea Finn
@chelseabfinn
I’m thrilled and honored to have received the MIT TR 35 innovator award this year. Grateful to have been able to work with some amazing faculty, students, and researchers thus far in my career!
@techreview #35innovatorsunder35 www.technologyreview.com/lists/innovators-under-35/2018/p… -
Ian Goodfellow reposted
WiCV
@WiCVworkshop
Call for submissions! The 5th Women in Computer Vision Workshop (WiCV) at #eccv2018 invites all vision researchers to submit their work and attend to our workshop. More info: wicvworkshop.github.io/ECCV2018
-
Ian Goodfellow reposted
Negar Rostamzadeh
@negar_rz
Excited to announce the FashionGen Text to Image challenge. The task is to generate fashion items based on design descriptions. Data is annotated by designers, with multiple angles of each item and has over 300,000 samples! @SSENSE @element_ai
fashion-gen.com -
Ian Goodfellow reposted
Communications of the ACM
@CACMmag
"Making Machine Learning Robust Against Adversarial Inputs," by @goodfellow_ian et al., looks at protecting #machinelearning systems against adversarial inputs. #ai cacm.acm.org/magazines/2018/7/229030-making-machine-learn…
-
Ian Goodfellow reposted
Black in AI
@black_in_ai
Calling all allies! We're setting up a mentoring program and would love to have you all involved.
If you work in AI (defined broadly, form below) could you please fill out this form? You can be in any stage of your career.
docs.google.com/forms/d/e/1FAIpQLSd85BdKTNiN2ZJnMJPI37dli… -
Ian Goodfellow
@goodfellow_ian
Alyosha Efros on “reasons to love GANs” at the #cvpr2018 tutorial on GANs
-
Ian Goodfellow
@goodfellow_ian
Carl Vondrick on video generation at the #cvpr2018 GAN tutorial
-
Ian Goodfellow reposted
Adriana Romero-Soriano
@adri_romsor
Carol E. Reiley from @driveai_ joining us remotely for her keynote talk at #wicv18 #cvpr18.
-
Ian Goodfellow
@goodfellow_ian
I'll present a talk called "Defense Against the Dark Arts" summarizing the state of the art and key research challenges for defenses against adversarial examples. Room 259, 1:30 PM. www.iangoodfellow.com/slides/2018-06-22-cv-cops.pdf
-
Ian Goodfellow
@goodfellow_ian
Sanjeev Arora on the theory of finite GANs at the #cvpr2018 tutorial
-
Ian Goodfellow
@goodfellow_ian
Taesung Park on CycleGAN at the #cvpr2018 GAN tutorial
-
Ian Goodfellow
@goodfellow_ian
Jun-Yan Zhu teaches how to do unsupervised image to image translation at the #cvpr2018 tutorial on GANs
-
Ian Goodfellow
@goodfellow_ian
Phillip Isola on pix2pix at #cvpr2018 GAN tutorial
-
Ian Goodfellow
@goodfellow_ian
#cvpr2018 Check out the all-day tutorial on GANs tomorrow: sites.google.com/corp/view/cvpr2018tutorialongans/ I'll speak at 9AM giving an introduction to GANs. Many great speakers throughout the day!
-
Ian Goodfellow
@goodfellow_ian
Consider submitting papers to ALEC, the symposium on Adversary-Aware Learning Techniques and Trends in Cybersecurity sites.google.com/corp/view/alec18
-
Ian Goodfellow reposted
Depth First Learning
@DepthFirstLearn
We just released our newest study guide! Learn all about TRPO from professors @kumarkagrawal and @suryabhupa → depthfirstlearning.com/2018/TRPO.
-
Ian Goodfellow reposted
Bryan Catanzaro
@ctnzr
.@tcwang0509 will be presenting an oral presentation Thursday at CVPR on his Pix2PixHD work generating high-resolution synthetic images with conditional GANs. This is a big step towards a generative model for rendering graphics. news.developer.nvidia.com/generating-and-editing-high-res…
-
Ian Goodfellow reposted
Google AI
@GoogleAI
Attending #cvpr2018 this year? Learn more about the computer vision research (and demos!) we are presenting this week in Salt Lake City, Utah. blog.research.google/2018/06/google-at-cvpr-2018.html
-
Ian Goodfellow
@goodfellow_ian
#cvpr2018 I will teach an Introduction to GANs at 8:45 AM in Room 150-ABC at the Perception Beyond the Visible Spectrum workshop. Slides available at www.iangoodfellow.com/slides/
-
Ian Goodfellow
@goodfellow_ian
Consider submitting papers to the Workshop on Positive and Negative Adversaries in Multimedia: project.inria.fr/panamm2018/
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
We're continuing to expand our @GoogleAI teams around the world. We’ll be opening our first research center in Africa in #ghana later this year! blog.google/intl/en-africa/company-news/technology/google…
If you're a machine learning researcher interested in working in Accra, Ghana, apply: www.google.com/about/careers/applications/jobs/results -
Ian Goodfellow reposted
Alex Mordvintsev
@zzznah
-
Ian Goodfellow reposted
Sundar Pichai
@sundarpichai
Today we’re sharing our AI principles and practices. How AI is developed and used will have a significant impact on society for many years to come. We feel a deep responsibility to get this right. blog.google/topics/ai/ai-principles/
-
Ian Goodfellow reposted
Dumitru Erhan
@dumierhan
AI at Google: our principles: blog.google/topics/ai/ai-principles/. Happy to see this out!
-
Ian Goodfellow reposted
Yaroslav Ganin
@yaroslav_ganin
Just learned that our paper "Synthesizing Programs for Images using Reinforced Adversarial Learning" (aka SPIRAL) got a long talk at @icmlconf. If you haven't already, go check it out: arxiv.org/abs/1804.01118
-
Ian Goodfellow reposted
Depth First Learning
@DepthFirstLearn
Announcing depthfirstlearning.com! We are building a repository of study guides targeting consequential papers. Check it out, learn something in-depth, and help us build the next one. @avitaloliver @suryabhupa @kumarkagrawal @cinjoncin
-
Ian Goodfellow reposted
Ben Recht
@beenwrekt
Do CIFAR-10 classifiers generalize to CIFAR-10? The answer will surprise you! arxiv.org/abs/1806.00451
-
Ian Goodfellow reposted
GeekPwn
@GeekPwn
Registration for Competition on Adversarial Attacks & Defences(CAAD)is open now! caad.geekpwn.org
-
Ian Goodfellow
@goodfellow_ian
These amazing students are working hard to make sure that AI is developed safely and responsibly
Coefficient Giving
@coeff_giving
Excited to announce our first class of AI Fellows, seven machine learning students to whom we’re collectively recommending $1.1 million in PhD fellowship support over the next five years: www.openphilanthropy.org/focus/global-catastrophic-risks/…
-
Ian Goodfellow
@goodfellow_ian
Two years of GAN progress on class-conditional ImageNet-128
-
Ian Goodfellow reposted
niki parmar
@nikiparmar09
tl;dr Check out our new results on Non Autoregressive MT!! We come very close to a greedy Transformer baseline while being >3x faster!
Arvind Neelakantan
@arvind_io
We develop a non-autoregressive machine translation model whose accuracy almost matches a strong greedy autoregressive baseline Transformer, while being 3.3 times faster at inference. Joint work with @ashVaswani @nikiparmar09 Aurko Roy arxiv.org/abs/1805.11063
-
Ian Goodfellow
@goodfellow_ian
Forbes listed GANs as one of the best tech innovations of the last three years: www.forbes.com/sites/forbestechcouncil/2018/05/25/the-bes…
-
Ian Goodfellow
@goodfellow_ian
I'm speaking at the 1st Deep Learning and Security workshop (co-located with @IEEESSP ) at 1:30 today: www.ieee-security.org/TC/SPW2018/DLS/ I'll discuss research into defenses against adversarial examples, including future directions. Slides and lecture notes here: www.iangoodfellow.com/slides/2018-05-24-DLS-notes.pdf
-
Ian Goodfellow
@goodfellow_ian
Self-attention for GANs. No more problems with losing track of how many faces the generator has drawn on the dog.
-
Ian Goodfellow
@goodfellow_ian
This paper shows how to make adversarial examples with GANs. No need for a norm ball constraint. They look unperturbed to a human observer but break a model trained to resist large perturbations. arxiv.org/pdf/1805.07894.pdf
-
Ian Goodfellow
@goodfellow_ian
Interpolating through the latent space of ramen dishes uses a GAN
Kenji Doi
@knjcode
Google Driveにラーメン二郎GAN動画の元ファイルを置きました。 drive.google.com/open?id=16w8k40RfKhGLqTL1nvqGgB_P19lT-I7v #ラーメン二郎 #gan
-
Ian Goodfellow
@goodfellow_ian
GANs can be used to automatically design dental crowns, that are then actually manufactured and used in the physical world. Crowns need to be made specifically for each patient and need to fit correctly with the other teeth and support biting and chewing. arxiv.org/abs/1804.00064
-
Ian Goodfellow reposted
Google AI
@GoogleAI
Congratulations to the first set of fellows from the Google Africa PhD Fellowship program! We are committed to advancing cutting-edge computer science research and building strong relationships with the academic community in Africa.
Google in Africa
@googleafrica
We are excited to announce the first 4 recipients of the Google Africa PhD Fellowships created to support early-stage PhD students enrolled in any African university in computer science and related fields. blog.google/intl/en-africa/company-news/inside-google/ann…
-
Ian Goodfellow
@goodfellow_ian
GANs for generating Mario levels! www.fastcompany.com/40568981/the-machines-have-taught-the…
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Nice write-up by Yonghui Wu of the Google Brain team about how the new Smart Compose feature in Gmail works behind-the-scenes.
Google AI
@GoogleAI
Learn more about the research behind Smart Compose, a new feature in Gmail that uses machine learning to interactively offer sentence completion suggestions as you type, allowing you to draft emails faster → research.google/blog/smart-compose-using-neural-networks-…
-
Ian Goodfellow reposted
Wojciech Zaremba
@woj_zaremba
Moore like law for results in AI: blog.openai.com/ai-and-compute/
-
Ian Goodfellow reposted
-
Ian Goodfellow
@goodfellow_ian
A math trick I like a lot is the approach to taking derivatives using hyperreal numbers. Thread:
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
There are also infinite hyperreals, larger than any real number, but I don't personally find them quite as useful for the math I need to do for machine learning research.
Ian Goodfellow
@goodfellow_ian
It is also possible to compute integrals using hyperreal numbers, but I don't personally find as much of an advantage to hyperreal numbers over limits for integration as I do in the case of derivation.
-
Ian Goodfellow
@goodfellow_ian
A 1904 Sherlock Holmes story predicted that if a great war between European powers came to pass it would result in 100,000 deaths. In reality, WW1 began 10 years later, and resulted in over 10 million deaths.
-
Ian Goodfellow
@goodfellow_ian
A quick thread on two of my favorite theory hacks for machine learning research
Ian Goodfellow
@goodfellow_ian
A lot of the time, we want to analyze the optimal behavior of a neural net using algebra / calculus. Neural net models are usually too complicated for you to algebraically solve for the parameters that optimize most functions (unless it's some trivial function like weight decay)
Ian Goodfellow
@goodfellow_ian
To get a less complicated model, a common instinct is to use a linear model. This is nice because it makes a lot of optimization problems convex. But it has a downside: a linear model can't do a lot of things a neural net can do. The solution becomes very oversimplified.
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
My co-authors and I used both theory hack #1 and #2 to derive eq 2 of the GAN paper: papers.nips.cc/paper/5423-generative-adversarial-nets.pdf
Ian Goodfellow
@goodfellow_ian
Bonus: A great source of related theory hacks is Sec 3.2 of web.stanford.edu/~boyd/cvxbook/bv_cvxbook.pdf . You can use these properties to prove that functions of functions are convex by construction using composition rules. No need to prove that second functional derivatives are positive, etc.
-
Ian Goodfellow reposted
niki parmar
@nikiparmar09
Our paper, Image Transformer got accepted to ICML! 💃🏻💃🏻💃🏻
niki parmar
@nikiparmar09
Our new paper “Image Transformer”, extends self-attention from the original Transformer to much longer sequences on Image Generation and Super-Resolution tasks. It beats previous SOTA autoregressive models like PixelCNN, PixelRNN. arxiv.org/abs/1802.05751
-
Ian Goodfellow reposted
Koichi Hamada
@hamadakoichi
Full-body High-resolution Anime Generation with Progressive Structure-conditional GANs.
- Structural conditions on both the generator and discriminator at each scale.
- Plausible HR image generation (512x512) consisting with the structures.
Details: dena.com/intl/anime-generation -
Ian Goodfellow reposted
tom white
@dribnet
-
Ian Goodfellow reposted
Stanislav Nikolov
@snikolov
okay this has probably been done but
-
Ian Goodfellow reposted
Lucas Theis
@lucastheis
Thought-provoking paper showing that high perceptual quality is incompatible with low distortion. Counter to intuition, this is true for any full-reference metric including perceptual metrics.
arxiv.org/abs/1711.06077 -
Ian Goodfellow reposted
DrVeronikaCH@dair-community.social
@DrVeronikaCH
If you are getting an error while trying to read the How I Fail post by @goodfellow_ian, this URL works
zubairahmed_ai
@zubairahmed_ai
Anyone who is following this, here's the working link while this blog restores web.archive.org/web/20180505190115/http://www.veronikach.…
-
Ian Goodfellow reposted
DrVeronikaCH@dair-community.social
@DrVeronikaCH
How I Fail - Ian Goodfellow (PhD'13, Computer Science) - interview with @goodfellow_ian www.veronikach.com/how-i-fail/how-i-fail-ian-goodfellow-p…
-
Ian Goodfellow reposted
Yasmin Green
@yasmind
We're hiring! If you are/know someone who enjoys ambiguity, research and technology apply! (And you can maker the world a safer place!)
Jigsaw
@Jigsaw
We're hiring a research program manager to investigate opportunities for technology to help vulnerable people around the world. careers.google.com/jobs&
-
Ian Goodfellow reposted
Andrew M. Dai
@AndrewDai
Come see our poster on using gradient penalties for general GANs being presented right now at ICLR.
-
Ian Goodfellow reposted
Jun-Yan Zhu
@junyanz89
The tutorial will be on Friday, June 22nd. In addition to the organizers, we also have a lineup of awesome speakers who will cover topics ranging from computer vision to robotic learning.
Ian Goodfellow
@goodfellow_ian
If you will attend #cvpr2018 , check out the tutorial on GANs: sites.google.com/corp/view/cvpr2018tutorialongans/ (organized by @junyanz89, Taesung Park, @elaClaudia , @phillip_isola and me)
-
Ian Goodfellow reposted
Anna Goldie
@annadgoldie
Come visit @Azaliamirh, @JeffDean, and my ICLR poster today from 11am-1pm, Poster #4!
-
Ian Goodfellow
@goodfellow_ian
If you will attend #cvpr2018 , check out the tutorial on GANs: sites.google.com/corp/view/cvpr2018tutorialongans/ (organized by @junyanz89, Taesung Park, @elaClaudia , @phillip_isola and me)
-
Ian Goodfellow reposted
Ari Morcos
@arimorcos
Thought-provoking work from @julius_adebayo, @goodfellow_ian and others showing that saliency maps don't change much, even when *all* the weights are reinitialized! Do saliency maps tell us what the network cares about, or merely what the task demands? openreview.net/forum?id=SJOYTK1vM
-
Ian Goodfellow reposted
Cody Coleman
@codyaustun
The first iteration of DAWNBench is done! We saw a 477x improvement in ImageNet training time and a 23x improvement in cost! dawn.cs.stanford.edu/2018/04/30/dawnbench-v1-results/
-
Ian Goodfellow reposted
Avital Oliver
@avitaloliver
Come say hi at ICLR!
We're presenting our work on realistic evaluation of semi-supervised learning algorithms at the workshop poster session on Monday at 4:30pm. (arxiv.org/abs/1804.09170) -
Ian Goodfellow reposted
David Sussillo
@SussilloDavid
Google Cloud Platform announces new credits program for researchers
blog.google/topics/google-cloud/google-cloud-platform-ann… -
Ian Goodfellow reposted
Mihaela Rosca
@MihaelaCRosca
Check out our poster at ICLR showcasing our work on GANs on Tuesday. @LiamFedus @balajiln @shakir_za @goodfellow_ian
Google DeepMind
@GoogleDeepMind
Details of all our #iclr2018 papers in one place -deepmind.com/blog/deepmind-papers-iclr-2018/
-
Ian Goodfellow reposted
hardmaru
@hardmaru
Using GANs and CMA-ES to procedurally generate Mario levels 🎮
Sebastian Risi
@risi1979
New @GECCO2018 paper with @CIGbalance, Schrum, @_JialinLiu, @squashstar, @rndmcnlly: "Evolving Mario Levels in the Latent Space of a DCGAN". We can create levels of increasing difficulty by evolving a sequence of latent GAN vectors.
PDF: goo.gl/7vwJrD -
Ian Goodfellow reposted
Jenn Wortman Vaughan
@jennwvaughan
Ok, machine learning people, what are the topics you would be most excited to see covered in a #nips2018 tutorial?
-
Ian Goodfellow reposted
@timnitGebru (@dair-community.social/bsky.social)
@timnitGebru
We need many more women applying to be TAs at Addis Coder this summer: www.addiscoder.com 50% of our students will be women. Please contact me if you have any questions, or send me a private message. @WiDS_Conference @wimlds it would be great to get help spreading the word.
-
Ian Goodfellow reposted
Delip Rao
@deliprao
Semi supervised learning makes it into Amazon shareholders letter by @JeffBezos www.sec.gov/Archives/edgar/data/1018724/00011931251812116…
-
Ian Goodfellow
@goodfellow_ian
A common theme in ML research the past few years: when we move past supervised learning for labeling the test set, evaluation is one of the hardest parts of the research. Comes up in generative models, security, RL, interpretability, FAT, etc
Ian Goodfellow
@goodfellow_ian
I personally agree with this. It’s the metric I started using in arxiv.org/abs/1412.6572. If you want to handle varying eps just show an eps-vs-success rate curve. But more generally we should develop better metrics than norm balls
Battista Biggio
@biggiobattista
I love the idea of a challenge, but we should stop considering min-distance attacks and switch to max-confidence ones, i.e., attacks in which you maximize the prob of evasion under a constrained perturbation |x'-x| < eps. These are more powerful and better measure DNN security
Bethge Lab
@bethgelab
We are happy to announce the #nips2018 #adversarialvisionchallenge! It's designed to inspire progress towards more robust vision models & real-world attacks. Great to team up with @googleresearch, @crowd_ai and @pennstate & to have @awscloud as a sponsor! www.crowdai.org/challenges/nips-2018-adversarial-vision-c…
-
Ian Goodfellow
@goodfellow_ian
I personally agree with this. It’s the metric I started using in arxiv.org/abs/1412.6572. If you want to handle varying eps just show an eps-vs-success rate curve. But more generally we should develop better metrics than norm balls
Battista Biggio
@biggiobattista
I love the idea of a challenge, but we should stop considering min-distance attacks and switch to max-confidence ones, i.e., attacks in which you maximize the prob of evasion under a constrained perturbation |x'-x| < eps. These are more powerful and better measure DNN security
Bethge Lab
@bethgelab
We are happy to announce the #nips2018 #adversarialvisionchallenge! It's designed to inspire progress towards more robust vision models & real-world attacks. Great to team up with @googleresearch, @crowd_ai and @pennstate & to have @awscloud as a sponsor! www.crowdai.org/challenges/nips-2018-adversarial-vision-c…
-
Ian Goodfellow reposted
-
Ian Goodfellow reposted
Emma Pierson
@2plus2make5
I've been thinking about mental health/quality of life among PhD students recently. If you are or were a PhD student, please help me out by answering this short survey :) docs.google.com/forms/d/e/1FAIpQLSeKmv0r_OCDzoz9uDqbjlll4…
-
Ian Goodfellow reposted
Rachel Thomas
@math_rachel
1 in 3 heart surgeons refuse difficult operations to avoid poor mortality ratings, survey shows www.telegraph.co.uk/science/2016/06/03/one-in-three-heart… h/t The Tyranny of Metrics
-
Ian Goodfellow
@goodfellow_ian
Blog post about Google talks by @elie and me about ML and security at @RSAConference yesterday: security.googleblog.com/2018/04/leveraging-ai-to-protect-…
-
Ian Goodfellow reposted
Bethge Lab
@bethgelab
We are happy to announce the #nips2018 #adversarialvisionchallenge! It's designed to inspire progress towards more robust vision models & real-world attacks. Great to team up with @googleresearch, @crowd_ai and @pennstate & to have @awscloud as a sponsor! www.crowdai.org/challenges/nips-2018-adversarial-vision-c…
-
Ian Goodfellow reposted
Avinash Hindupur
@hindupuravinash
Exactly an year after The GAN Zoo was compiled.. research on GANs is still growing at an explosive pace! Here's a great thread by @goodfellow_ian on how to review papers and keep up-to-date!
Ian Goodfellow
@goodfellow_ian
Thread on how to review papers about generic improvements to GANs
-
Ian Goodfellow reposted
Elie Bursztein
@elie
How to successfully harness AI to combat fraud and abuse
www.elie.net/talk/how-to-successfully-harness-ai-to-comba… slides & video recording #ai #rsac2018 @googleresearch -
Ian Goodfellow
@goodfellow_ian
I’m speaking at @RSAConference at 3PM today about security and privacy of machine learning: www.rsaconference.com/speakers/ian-goodfellow
-
Ian Goodfellow reposted
Alexey Kurakin
@alexey2004
We released code of the backend which was used for evaluation of NIPS17 adversarial competition: github.com/tensorflow/cleverhans/tree/master/examples/nip…
-
Ian Goodfellow reposted
Ming-Yu Liu
@liu_mingyu
MUNIT: Multimodal unsupervised image-to-image translation. Learn to translate one input dog image to a distribution of cat images without paired training data.
paper: arxiv.org/abs/1804.04732
code: github.com/nvlabs/munit
by @xunhuang1995 @SergeBelongie @jankautz
#deeplearning -
Ian Goodfellow reposted
Jeff Dean
@JeffDean
I'm very much looking forward to being one of the speakers at the 2018 Indaba in South Africa in September. Applications close in a few weeks!
Deep Learning Indaba
@DeepIndaba
Applications to participate in the 2018 Indaba close in a few weeks. Encourage as many as you know to apply at deeplearningindaba.com/apply. We can't wait to learn from the inspirational speakers we have lined up. More details on this soon ...
-
Ian Goodfellow
@goodfellow_ian
This is one reason to do more work on adversarial examples in domains other than images. For images most points in R^n are hard to label for philosophical reasons. Could get around this by studying, e.g., adversarial lists of numbers that Neural Turing Machines sort incorrectly.
Catherine Olsson
@catherineols
A:... We define a "mistake" for an image classifier as disagreeing with the "true label" for the image.
So: what is the true label for a synthetic or altered image, not captured by a camera? Is it what a human would label it? Undefined, because no real object was photographed? -
Ian Goodfellow
@goodfellow_ian
The definition of "adversarial examples" I prefer these days is "Adversarial examples are inputs to machine learning models that an attacker has intentionally designed to cause the model to make a mistake" blog.openai.com/adversarial-example-research/
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
It isn't specific to classification; RL agents, generative models, autoencoders etc. all make different kinds of "mistakes" that aren't misclassification
Ian Goodfellow
@goodfellow_ian
The main downside of this definition: it's not a formal definition
-
Ian Goodfellow reposted
Julia Galef
@juliagalef
We're hiring! @open_phil is a small team deciding how to give away billions of $; working here is a very high-leverage way to impact humanity. If you're analytical, truth-seeking, & comfortable making EV estimates, you should join us. More about the job: medium.com/@acotra2017/hiring-analytical-thinkers-to-help…
-
Ian Goodfellow
@goodfellow_ian
Extreme image compression with GANs: data.vision.ee.ethz.ch/aeirikur/extremecompression/ (h/t @D_Berthelot_ML )
-
Ian Goodfellow
@goodfellow_ian
Now that the FCC is no longer providing net neutrality, the state of California can do so. If you support the state bill, you can say so here: actionnetwork.org/letters/add-your-name-save-net-neutrali…
-
Ian Goodfellow reposted
Prosthetic Knowledge
@prostheticknowl
@hardmaru @quasimondo @samim TempoGAN: volumetric GAN for Super Resolution Fluid Flow ge.in.tum.de/publications/tempogan/
-
Ian Goodfellow reposted
Greg Brockman
@gdb
Reinforcement learning today lets you solve a particular environment, without learning skills that transfer to new environments. We've just launched a contest and dataset where agents use their experience to adapt to new environments: blog.openai.com/retro-contest/
-
Ian Goodfellow reposted
Fermat's Library
@fermatslibrary
This Thursday we are doing an AMA with Augustus Odena (@gstsdn) from Google Brain about local geometry of GAN generators. You can ask Augustus questions about the paper here:
fermatslibrary.com/arxiv_comments?url=https%3A%2F%2Farxiv… -
Ian Goodfellow reposted
Alexey Kurakin
@alexey2004
Just released on arxiv a book chapter about "Adversarial attacks and defenses competition" which was part of NIPS17 competition track. arxiv.org/abs/1804.00097
-
Ian Goodfellow reposted
Karol Kurach
@karol_kurach
MemGEN paper released - generative modeling finally solved with @sylvain_gelly.
GANs are cool @goodfellow_ian but #deepmemory will completely revolutionize the Machine Learning field.
arxiv.org/abs/1803.11203 -
Ian Goodfellow
@goodfellow_ian
-
Ian Goodfellow reposted
Google DeepMind
@GoogleDeepMind
By learning to write programs that generate images our artificial agents can reason about how digits, characters and portraits are constructed. Read the blog: deepmind.com/blog/learning-to-generate-images/
-
Ian Goodfellow reposted
nature
@Nature
PhD and master’s students worldwide report rates of depression and anxiety that are six times higher than those in the general public. What do you think should be done to help? go.nature.com/2Gd3ady
-
Ian Goodfellow
@goodfellow_ian
Someone e-mailed me to say "The link between nipples and NIPS happened in your mind and was purely vindictive." Search engines seem to think there's a connection.
-
Ian Goodfellow reposted
-
Ian Goodfellow reposted
Kristian Lum
@KLdivergence
Seeing my essay cited as a reason that prominent people in statistics and ML are taking a stand for a more inclusive conference culture makes the stress of the last several months all worth it.
-
Ian Goodfellow
@goodfellow_ian
NIPS has been, on at least one occasion, a very stressful environment for me, and when I tried to get help via the existing code of conduct it did more harm than good. I think many ML people perceive me as invulnerable, so imagine what it's like to be a junior grad student.
Ian Goodfellow
@goodfellow_ian
I personally would also very much like to see NIPS 1) create and enforce a stronger code of conduct, and 2) change its name to something less prone to sexual puns and NSFW search results.
-
Ian Goodfellow
@goodfellow_ian
I personally would also very much like to see NIPS 1) create and enforce a stronger code of conduct, and 2) change its name to something less prone to sexual puns and NSFW search results.
-
Ian Goodfellow
@goodfellow_ian
In the first paper about adversarial examples, we focused on imperceptible perturbations. Since 2014, we've worked on defending against larger perturbations. Today, ALP focuses on clearly perceptible perturbations that occupy about 1/8 the dynamic range of the pixels.
-
Ian Goodfellow
@goodfellow_ian
Answering AMA questions now
Ian Goodfellow
@goodfellow_ian
.@harinidkannan , @alexey2004 and I will do an AMA on Adversarial Logit Pairing, the state of the art defense against adversarial examples on ImageNet, via @fermatslibrary on Tuesday Mar 27: fermatslibrary.com/arxiv_comments?url=https%3A%2F%2Farxiv…
-
Ian Goodfellow reposted
Oriol Vinyals
@OriolVinyalsML
RL + GANs: Program synthesis with an agent that uses a paint program to fool a discriminator. Paper+Blog: deepmind.com/blog/learning-to-generate-images/
-
Ian Goodfellow
@goodfellow_ian
Hadrien Jean has made a series of notes on linear algebra, following the deep learning textbook: hadrienj.github.io/posts/Deep-Learning-Book-Series-Introd…
-
Ian Goodfellow
@goodfellow_ian
I'm frustrated with how much conflict / unjustified demand for credit there is in deep learning today. Someone wrote to one of the Brain Residents I mentor asking to be cited for an idea that they say they "clumsily thought of then neglected trying it"
-
Ian Goodfellow
@goodfellow_ian
2nd thread on evaluating GAN papers (1st thread hit max thread length)
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
Mode collapse can also be very subtle repetition of textures or backgrounds in images that otherwise might look diverse to the human eye.
Ian Goodfellow
@goodfellow_ian
That's about it for today's brain-dump. Feel free to reply with your own advice for GAN paper reviewers.
-
Ian Goodfellow
@goodfellow_ian
Thread on how to review papers about generic improvements to GANs
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
My #1 recommendation is that reviewers of GAN papers should read "Are GANs Created Equal?" ( arxiv.org/pdf/1711.10337.pdf ) for an explanation of why empirical work in this area is hard and how to do it right
Ian Goodfellow
@goodfellow_ian
Another good paper to read for background is "A note on the evaluation of generative models" ( arxiv.org/abs/1511.01844 ) which explains why it is possible to have models with great samples and bad likelihood or vice versa and other issues with metrics for generative models.
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
It is extremely important to explain where all hyperparameters came from. Often new methods just seem like improvements because the authors spent more time informally optimizing the hyperparameters for the new method
Ian Goodfellow
@goodfellow_ian
Achievement unlocked: max twitter thread length. I will continue in another thread
-
Ian Goodfellow
@goodfellow_ian
1/11) Thread on bidding to review conference papers
Post not available here, view on X
Ian Goodfellow
@goodfellow_ian
10/11) Automatic reviewer assignment systems often have to ignore a lot of bids to make sure every paper gets reviewers. I've definitely been assigned to papers where I bid "not willing" before and several of my friends tell me they have too.
Ian Goodfellow
@goodfellow_ian
11/11) If I'm your area chair, you'll almost certainly get one reviewer who is an absolute expert on the topic of your paper. I even recruit new reviewers when I need to. For example, I often recruit reviewers from the security community to review ML security papers.
-
Ian Goodfellow
@goodfellow_ian
GANs for designing DNA: arxiv.org/pdf/1712.06148.pdf
-
Ian Goodfellow reposted
Catherine Olsson
@catherineols
I'm co-organizing an ICML workshop to host debates on the future of AI: www.machinelearningdebates.com
Notably, the focus is *not* on smack-downs and controversy; rather on making space for nuance, discussing falsifiable predictions, and changing your own and other's minds. [1/2] -
Ian Goodfellow
@goodfellow_ian
.@harinidkannan , @alexey2004 and I will do an AMA on Adversarial Logit Pairing, the state of the art defense against adversarial examples on ImageNet, via @fermatslibrary on Tuesday Mar 27: fermatslibrary.com/arxiv_comments?url=https%3A%2F%2Farxiv…
-
Ian Goodfellow
@goodfellow_ian
Check out Adversarial Logit Pairing, the new state of the art defense against adversarial examples on ImageNet, by @harinidkannan @alexey2004 and I: arxiv.org/abs/1803.06373
-
Ian Goodfellow reposted
The TWIML AI Podcast
@twimlai
Today we're joined by @goodfellow_ian of @GoogleBrain @googleresearch and Sandy Huang, PhD Student at @UCBerkeley, to discuss their paper "Adversarial Attacks on Neural Network Policies." Follow the link to check it out! twimlai.com/go/119
-
Ian Goodfellow
@goodfellow_ian
I'm doing an AMA with @fermatslibrary today: fermatslibrary.com/arxiv_comments?url=https%3A%2F%2Farxiv…
-
Ian Goodfellow reposted
Matt Lieberman
@social_brains
Best illusion I've seen, maybe ever. The blue and black lines are NOT moving. Insanely good.
-
Ian Goodfellow reposted
Fermat's Library
@fermatslibrary
We are doing an AMA (Ask Me Anything) about the GANs (Generative Adversarial Nets) paper with @goodfellow_ian on 3/15 (Thursday).
You can already ask questions to Ian here: fermatslibrary.com/arxiv_comments?url=https%3A%2F%2Farxiv… -
Ian Goodfellow
@goodfellow_ian
Debate over the cat-dog during Blake Richards' talk at #cosyne (sent by @thisismyhat to all cat-dog co-authors) lasted long enough that the projector overheated
-
Ian Goodfellow reposted
Black in AI
@black_in_ai
"We’re providing 6-10 stipends and mentorship to individuals from underrepresented groups to study deep learning full-time for 3 months and open-source a project" OpenAI
Check out the new Open AI Scholars program and spread the word ! blog.openai.com/openai-scholars/ -
Ian Goodfellow reposted
-
Ian Goodfellow
@goodfellow_ian
4 years of GAN progress (source: www.eff.org/files/2018/02/20/malicious_ai_report_final.pdf )
-
Ian Goodfellow
@goodfellow_ian
Yoshua, Aaron, and I have released the LaTeX template for the Deep Learning book: github.com/goodfeli/dlbook_notation Useful if you want to follow the same math notation conventions as we do or if you want to put a notation page in your document
-
Ian Goodfellow
@goodfellow_ian
Adversarial examples for humans in IEEE Spectrum: spectrum.ieee.org/the-human-os/robotics/artificial-intell…
-
Ian Goodfellow
@goodfellow_ian
Some press coverage of adversarial examples for humans: thenextweb.com/google/2018/02/28/google-teaches-ai-to-foo…
-
Ian Goodfellow
@goodfellow_ian
And the paper on real-world adversarial masks: www.cs.cmu.edu/~sbhagava/papers/face-rec-ccs16.pdf
Catherine Olsson
@catherineols
"They were all wearing adversarial masks [...] our object detectors told our security system that 'three chairs are running at 15 kilometers per hour down the corridor'"
More delightful fiction from @jackclarkSF's Tech Tales
jack-clark.net/2018/02/26/import-ai-83-cloning-voices-wit… -
Ian Goodfellow
@goodfellow_ian
By looking at this image, you can see how sensitive your own eyes are to contrast at different frequencies (taller apparent peaks=more sensitivity at that frequency). It's like a graph that is made by perceiving the graph itself. h/t @catherineols fourier.eng.hmc.edu/e180/lectures/v1/node11.html
-
Ian Goodfellow
@goodfellow_ian
Adversarial examples that fool both human and computer vision arxiv.org/abs/1802.08195
-
Ian Goodfellow
@goodfellow_ian
MIT Technology Review listed GANs as 1 of 10 breakthrough technologies this year: www.technologyreview.com/lists/technologies/2018/
-
Ian Goodfellow
@goodfellow_ian
A great Scientific American article on creativity and problem solving in machine learning by Chris Baraniuk: www.scientificamerican.com/article/for-ai-to-get-creative…
-
Ian Goodfellow
@goodfellow_ian
One of my main concerns about machine learning interpretability tools is that they will make people think they understand ML when they don't. People seem to think linear models are interpretable, but no one looks at them and has the intuition that they have adversarial examples
-
Ian Goodfellow
@goodfellow_ian
The term “deep learning” reminds me of “horseless carriage.” It made sense when introduced, but now that it is the dominant paradigm, it feels quaint to specify that there is no horse. The horse here is of course the shallow model / convex cost constraint.
-
Ian Goodfellow reposted
Yaroslav Bulatov
@yaroslavvb
Posted some time/cost evals of training image classifiers to dawn.cs.stanford.edu/benchmark/ . Somewhat counter-intuitively, the most expensive GPU (V100) ended up being the cheapest for training because of reduced training time
-
Ian Goodfellow reposted
Google AI
@GoogleAI
Congrats to Googlers Martín Abadi and Diane Greene, who have been elected to the National Academy of Engineering (NAE) for "contributions to the formal theory of computer security" and "contributions in transforming virtualization from a concept to an industry," respectively.
National Academy of Engineering
@theNAEng
OFFICIAL: The National Academy of Engineering announces the Class of 2018! We are proud to welcome 83 new members and 16 new foreign members. Congratulations to these incredible #engineers! Read the official press release: www.nae.edu/CheckRobotPage.html?redirectUrl=/MediaRoom/20…
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Congrats to my Google colleagues Martin Abadi ("For contrib. to the formal theory of computer security") and Diane Greene ("For contrib. in transforming virtualization from a concept to an industry") on their NAE election. This is one of the highest honors for an engineer!
National Academy of Engineering
@theNAEng
OFFICIAL: The National Academy of Engineering announces the Class of 2018! We are proud to welcome 83 new members and 16 new foreign members. Congratulations to these incredible #engineers! Read the official press release: www.nae.edu/CheckRobotPage.html?redirectUrl=/MediaRoom/20…
-
Ian Goodfellow
@goodfellow_ian
A nice plot of baselines conveniently worsening each year when it's time to write a new paper. e.humanities.uva.nl/ireval/papers/paper_2.pdf h/t @neal_lathia
-
Ian Goodfellow
@goodfellow_ian
I did a small scale version of this for one GAN metric. Paper with best baseline was rejected from ICLR 2018, papers with worse baselines were accepted to ICLR 2017 and NIPS 2017.
Ian Goodfellow
@goodfellow_ian
Meta-research idea: find all ICLR submissions that report baseline numbers for the same task. Determine whether higher baseline numbers are correlated with acceptance (due to better empirical work) or rejection (due to less perceived improvement).
-
Ian Goodfellow
@goodfellow_ian
Meta-research idea: find all ICLR submissions that report baseline numbers for the same task. Determine whether higher baseline numbers are correlated with acceptance (due to better empirical work) or rejection (due to less perceived improvement).
-
Ian Goodfellow
@goodfellow_ian
Performance of new method & performance of baseline was correlated across 10 ICLR papers. Positive interpretation: both new method and baseline perform proportional to skill, resources, etc. Cynical interpretation: papers w/weak results sandbagged the baseline. h/t @colinraffel
-
Ian Goodfellow reposted
Deborah Hanus
@deborahhanus
I've had several people ask me how the can start learning #machinelearning, #deeplearning, & #reinforcementlearning. I've compiled a list of some of my favorite resources for getting started. deborahhanus.com/blog/learn-machine-learning/
cc @fastdotai @goodfellow_ian -
Ian Goodfellow
@goodfellow_ian
I've seen several people recently say that instead of deep learning, AI researchers should reverse engineer the brain. This is strange to me, because that was one of the motivations for deep learning (and the name of the team, Google Brain)
Ian Goodfellow
@goodfellow_ian
Deep learning isn't intended to model brain function with high fidelity (that's computational neuroscience) but it's meant to work in somewhat brain-like ways, especially compared to what came before: kernel machines, Bayes nets with table CPDs, hand-engineered features
-
Ian Goodfellow
@goodfellow_ian
Code for SN-GAN and the projection discriminator are now available: github.com/pfnet-research/sngan_projection
-
Ian Goodfellow
@goodfellow_ian
Our paper on adversarial spheres has been accepted to the ICLR workshops. We show that even nearly perfect classifiers can suffer badly from adversarial examples, calling for us to reformulate our strategy to achieving machine learning security guarantees. openreview.net/forum?id=SyUkxxZ0b
-
Ian Goodfellow
@goodfellow_ian
MaskGAN, our paper on generating text with GANs, was accepted to ICLR: openreview.net/forum?id=ByOExmWAb
-
Ian Goodfellow
@goodfellow_ian
Our paper on using thermometer codes to resist adversarial examples has been accepted to ICLR. This model was highly robust to the best white box attacks we were able to design. openreview.net/forum?id=S18Su--CW
-
Ian Goodfellow
@goodfellow_ian
Our paper on Ensemble Adversarial Training has been accepted to ICLR. This approach increases robustness of models in the black box setting and was used by all the top defense teams in the NIPS 2017 adversarial example competition. openreview.net/forum?id=rkZvSe-RZ
-
Ian Goodfellow
@goodfellow_ian
Our ICLR submission "Many paths to equilibrium" has been accepted. This submission corrects a few common misconceptions about how GANs learn---tl;dr GANs are not literally Jensen-Shannon divergence optimizers and can solve problems Jensen-Shannon can't openreview.net/challenge?redirect=%2Fforum%3Fid%3DByQpn1ZA-
-
Ian Goodfellow
@goodfellow_ian
Users of @Strava have complained about the app's limited / confusing privacy settings since as far back as 2014. I myself posted a complaint in 2017 ( support.strava.com/hc/en-us/community/posts/208837267-Pri… ). Now the app has inadvertently revealed the location of secret military bases.
Carl Carrie (@🏠)
@carlcarrie
Fitness tracking app #strava gives away locations of secret military basis - #cybersecurity gap www.theguardian.com/world/2018/jan/28/fitness-tracking-ap…
-
Ian Goodfellow reposted
TensorFlow
@TensorFlow
TensorFlow 1.5.0 has been released, including previews of Eager execution (github.com/tensorflow/tensorflow/tree/r1.5/tensorflow/con…) and TensorFlow Lite (github.com/tensorflow/tensorflow/tree/r1.5/tensorflow/con…)!
Please see the full release notes for details on added features and changes ↓
github.com/tensorflow/tensorflow/releases/tag/v1.5.0 -
Ian Goodfellow reposted
Dumitru Erhan
@dumierhan
I'll be giving a talk on video prediction at #reworkdl this afternoon: www.re-work.co/events/deep-learning-summit-san-francisco-…
-
Ian Goodfellow reposted
Katie #reworkAI
@reworkkatie
Tomorrow, @goodfellow_ian will be signing his book "DEEP LEARNING" after his talk at the Deep Learning Summit. Make sure to pop by! #reworkdl #reworkai #deeplearning #googebrain
-
Ian Goodfellow
@goodfellow_ian
A correction: this video makes use of both SN-GAN and the technique from another paper, the projection discriminator: openreview.net/forum?id=ByS1VpgRZ
Ian Goodfellow
@goodfellow_ian
Video of morphing between ImageNet categories with SN-GAN: www.youtube.com/watch?v=r6zZPn-6dPY
-
Ian Goodfellow
@goodfellow_ian
Adversarial patch in @sciam : www.scientificamerican.com/article/how-to-hack-an-intelli…
-
Ian Goodfellow
@goodfellow_ian
My family and I certainly did not benefit from any of these supposed policy changes. I was separated from my wife and 1 year old baby for a transatlantic flight. My wife and baby were delayed by *4 days* and my wife had to handle the baby alone on this very long flight.
Ian Goodfellow
@goodfellow_ian
After the video of David Dao being violently dragged off a @united flight went viral, @united announced a new set of policies regarding overbooking: www.nbcnews.com/business/travel/united-aims-be-more-custo… I think that to avoid following these policies, they say there's a database glitch, not overbooking
Ian Goodfellow
@goodfellow_ian
In case anyone is wondering, this did not end well. I think @united has invented a new way to avoid paying people to volunteer not to get on the plane. They just prevent you from getting on and say it's a computer glitch.
Ian Goodfellow
@goodfellow_ian
Swiss Air says @united canceled my ticket (without notifying me in any way). The flight is still going and there is still space. I can buy a new ticket but they say I can’t get a refund for the canceled one.
-
Ian Goodfellow
@goodfellow_ian
After the video of David Dao being violently dragged off a @united flight went viral, @united announced a new set of policies regarding overbooking: www.nbcnews.com/business/travel/united-aims-be-more-custo… I think that to avoid following these policies, they say there's a database glitch, not overbooking
Ian Goodfellow
@goodfellow_ian
In case anyone is wondering, this did not end well. I think @united has invented a new way to avoid paying people to volunteer not to get on the plane. They just prevent you from getting on and say it's a computer glitch.
Ian Goodfellow
@goodfellow_ian
Swiss Air says @united canceled my ticket (without notifying me in any way). The flight is still going and there is still space. I can buy a new ticket but they say I can’t get a refund for the canceled one.
-
Ian Goodfellow
@goodfellow_ian
In case anyone is wondering, this did not end well. I think @united has invented a new way to avoid paying people to volunteer not to get on the plane. They just prevent you from getting on and say it's a computer glitch.
Ian Goodfellow
@goodfellow_ian
Swiss Air says @united canceled my ticket (without notifying me in any way). The flight is still going and there is still space. I can buy a new ticket but they say I can’t get a refund for the canceled one.
-
Ian Goodfellow
@goodfellow_ian
Video of morphing between ImageNet categories with SN-GAN: www.youtube.com/watch?v=r6zZPn-6dPY
-
Ian Goodfellow reposted
Lindsey
@lindsey
"Toward Scalable Verification for Safety-Critical Deep Networks" will appear at the inaugural SysML conference (sysml.cc) next month! See you at Stanford. arxiv.org/abs/1801.05950
-
Ian Goodfellow reposted
Y Combinator
@ycombinator
XIX.ai (W17) is organizing a conference about AI. You can get tickets here aimeetupseries.com
-
Ian Goodfellow reposted
raia hadsell
@RaiaHadsell
Announcing the inaugural Transylvanian Machine Learning Summer School! tmlss.ro Please pass on to eastern european students in ML and AI. I'll be one of the teachers, for robotics and deep RL.
-
Ian Goodfellow
@goodfellow_ian
My friend @yaroslavvb did a lot of work on this
OpenAI
@OpenAI
Releasing gradient checkpointing - a package for fitting bigger Tensorflow models onto your GPU. By trading off memory usage against a 20% increase in computation time, we’ve used this to train 10X larger models: github.com/openai/gradient-checkpointing
-
Ian Goodfellow
@goodfellow_ian
It’s strange to see people defining deep learning as supervised learning via backprop, considering that the 2006 deep learning revolution was originally based on the idea that neither of those things work very well
-
Ian Goodfellow
@goodfellow_ian
I will speak at this event
MIT Technology Review
@techreview
Meet the sharpest minds in AI & get the insight and intelligence you need to reinvent your business and your world at #emtechdigital. www.technologyreview.com/events?discount=SHARPMINDSTW
-
Ian Goodfellow
@goodfellow_ian
What the @GoogleBrain team accomplished in 2017: research.googleblog.com/2018/01/the-google-brain-team-loo…
-
Ian Goodfellow
@goodfellow_ian
Same goes for adversarial ML
David Sussillo
@SussilloDavid
For deep learning folks who really want to get down with recurrent neural networks, nonlinear dynamics should be on your study list.
This post is unavailable. See it on X. -
Ian Goodfellow reposted
Hillary Sanders
@hillarymsanders
Really awesome work done by old college friend Nicholas Carlini, showing that speech-to-text systems can be disturbingly successfully fooled by adversarial examples!
arxiv.org/pdf/1801.01944.pdf -
Ian Goodfellow
@goodfellow_ian
To gain some idea of the far future of ML security, we studied a simple toy problem called "adversarial spheres," simulating a future where advanced ML models are extremely accurate. We find that even then, an adversary can still easily fool them. arxiv.org/abs/1801.02774
-
Ian Goodfellow reposted
Kat Lo
@lolkat
To all those folks trying to automate online harassment detection and mitigation measures-- how carefully are you thinking about the ways you codify toxicity and harassment?
How are you validating the way you think toxicity and abuse affect community health and participation?
Pew Research Internet
@pewinternet
"Everyone has a different idea about what harassment is. Especially when race and gender are involved." @verge on our new report about online harassment www.theverge.com/2018/1/4/16846280/online-harassment-defi…
-
Ian Goodfellow
@goodfellow_ian
I never heard back from MIT. I got rejected from CMU. I was accepted to U of T but not to work with the PI I wanted there. I got "honorable mention" for NSF GRFP but not actual money. Don't let temporary failures discourage you.
-
Ian Goodfellow
@goodfellow_ian
Adversarial patch covered by the BBC: www.bbc.com/news/technology-42554735
-
Ian Goodfellow
@goodfellow_ian
The new Justin Timberlake video is at a deep learning conference. This is just surreal to me. www.youtube.com/watch?v=gA-NDZb29I4
-
Ian Goodfellow
@goodfellow_ian
An article about adversarial patch: www.fastcodesign.com/90156089/trippy-stickers-trick-compu…
-
Ian Goodfellow
@goodfellow_ian
Swiss Air says @united canceled my ticket (without notifying me in any way). The flight is still going and there is still space. I can buy a new ticket but they say I can’t get a refund for the canceled one.
-
Ian Goodfellow
@goodfellow_ian
I’ve been trying to check in to my @united flight operated by Swiss Air for an hour and 15 minutes. Not allowed to check in because “an agent touched” my reservation. I only have a few minutes left before it will be too late to get to my flight.
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Less than four days left to apply for Google's AI Residency program (formerly the Google Brain Residency program). Applications close on Monday, January 8th. If you want to do machine learning research, consider applying!
g.co/airesidency and
g.co/airesidency/apply -
Ian Goodfellow reposted
Rosemary Graves
@RosemaryHGraves
Help us find the young innovators shaping our future. Nominate a stellar technologist for our @techreview 35 Innovators Under 35.
This is a super cool award that dates back to the 1990's and has honored the likes of Zuck, Jack Dorsey and @AndrewYNg. www.technologyreview.com/s/609177/help-us-find-the-young-… -
Ian Goodfellow reposted
Toni Creswell
@ToniCreswell
Denoising Adversarial Autoencoders: Classifying Skin Lesions Using Limited Labelled Training Data arxiv.org/pdf/1801.00693.pdf.
-
Ian Goodfellow
@goodfellow_ian
ML paper writing pro-tip: you can download the raw source of any arxiv paper. Click on the "Other formats" link, then click "Download source". This gets you a .tar.gz with all the .tex files, all the image files for the figures in their original resolution, etc.
Ian Goodfellow
@goodfellow_ian
Suggestion for @arxiv_org : make this feature more prominent
-
Ian Goodfellow
@goodfellow_ian
An article about GANs in the @nytimes : www.nytimes.com/interactive/2018/01/02/technology/ai-gene…
-
Ian Goodfellow reposted
arxiv
@arxiv_org
Adversarial Patch. arxiv.org/abs/1712.09665
-
Ian Goodfellow reposted
Talk Python Podcast
@TalkPython
How is machine learning changing particle physics? Join @WonderMicky, @Michael_A_Kagan, and @HEPfeickert to learn how machine learning is being used at the large hadron collider on @talkpython talkpython.fm/144
-
Ian Goodfellow
@goodfellow_ian
I started my Google career as an intern designing new kinds of conv nets for Google Maps and had a great time there
Jeff Dean
@JeffDean
A really nice write-up (by someone outside Google) about the kinds of behind-the-scenes work that goes into taking lots of kinds of raw data and turning it all into high quality Google Maps.
www.justinobeirne.com/google-maps-moat -
Ian Goodfellow reposted
The TWIML AI Podcast
@twimlai
NEW SHOW ALERT!! @Samcharrington is joined by @TimnitGebru, postdoc at @MSFTResearch, and one of the organizers of @black_in_ai, to discuss her recent paper "Using Deep Learning and Google Street View to Estimate the Demographic Makeup of the US". twimlai.com/podcast/twimlai/using-deep-learning-and-googl…
-
Ian Goodfellow
@goodfellow_ian
DreamScapes 3, the latest Inceptionist artwork from Daniel Ambrosi: www.danielambrosi.com/Dreamscapes-3/
-
Ian Goodfellow
@goodfellow_ian
Learn about how to use Street View imagery for demographic research from @timnitGebruon on @twimlai
-
Ian Goodfellow reposted
tom white
@dribnet
thanks - prints for sale here; currently one fan print left.
dribnet.bigcartel.com -
Ian Goodfellow
@goodfellow_ian
Many thanks to @dribnet for giving me this print at #nips2017 Now proudly on the wall at home
-
Ian Goodfellow reposted
WiML
@WiMLworkshop
For over a decade, WiML has showcased the outstanding technical skills of female researchers. Today, we highlight that all researchers are *people* who deserve to be treated with dignity, respect and professionalism. There should be no tolerance for failing to meet this standard.
Kristian Lum
@KLdivergence
I'll probably end up remembering this moment as the time I decided to blow my career up for no reason, but this is something I've been wanting to say for a long time.
medium.com/@kristianlum/statistics-we-have-a-problem-3046… -
Ian Goodfellow reposted
Danielle Belgrave
@DaniCMBelg
Here's a copy of the slides from my presentation @black_in_ai workshop during #nips2017 "Machine Learning for Personalised Health": www.dropbox.com/scl/fi/ly5g6pizc3axwseqrc1yz/Danielle-Bel… . Thanks to @timnitGebru @red_abebe and organisers for putting together such an amazing workshop.
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Read the accounts of what Kristian Lum described, and if you see such behavior in your own workplace or conferences, speak up and make it clear that this is unacceptable. Thanks for describing your experiences, @KLdivergence.
Kristian Lum
@KLdivergence
I'll probably end up remembering this moment as the time I decided to blow my career up for no reason, but this is something I've been wanting to say for a long time.
medium.com/@kristianlum/statistics-we-have-a-problem-3046… -
Ian Goodfellow reposted
Miles Brundage
@Miles_Brundage
Reminder that @FHIOxford is still looking for interns in AI safety + RL (www.fhi.ox.ac.uk/fhi-is-accepting-applications-for-intern…) and policy/governance (www.fhi.ox.ac.uk/ai-policy-governance-internship/). Especially looking for candidates from underrepresented demographic groups, and people who are available at times other than summer.
-
Ian Goodfellow reposted
David Sussillo
@SussilloDavid
Folks, the Google AI Residency Program is accepting applications (due Jan 8).
So often I hear, "But I'm not a machine learning wiz." That is the whole point! We encourage a diverse set of backgrounds to apply! You do you. -
Ian Goodfellow
@goodfellow_ian
.@lawlkat is doing important work to make online communities more safe and fair: motherboard.vice.com/en_us/article/a3jkd4/the-reddit-mode…
-
Ian Goodfellow reposted
Google AI
@GoogleAI
In order to make Generative Adversarial Networks (GANs) easier to experiment with, we’ve open sourced TFGAN, a lightweight #tensorflow library designed to make it easy to train and evaluate GANs ➝ blog.research.google/2017/12/tfgan-lightweight-library-fo…
-
Ian Goodfellow reposted
MIT Technology Review
@techreview
“Through decades of public and private investment, the U.S. currently leads in basic AI research. Ill-advised policies can quickly squander this lead.” www.technologyreview.com/2017/12/06/3775/the-us-leads-in-…
-
Ian Goodfellow
@goodfellow_ian
I use this CMU font in all my slides so that pasted TeX matches the slide font. Same for figures in www.deeplearningbook.org
hardmaru
@hardmaru
Actually, @goodfellow_ian gave me a few pointers on fonts the other day. These CM-Unicode TrueType fonts were derived from the fonts created by Donald Knuth for TeX, and can be used in other applications. www.checkmyworking.com/cm-web-fonts/
-
Ian Goodfellow reposted
Jeff Dean
@JeffDean
Current U.S. policies that make it hard or impossible for some of the most talented people in the world to enter this country, or the proposed tax policies that will tax graduate students that make $28k per year as if they make $71k will definitely hurt the U.S.
Christopher Schroeder
@cmschroed
A good friend, AI scientist, recently returned from China in awe: “they are toe to toe with us, but moving faster.” In Singapore I was told, “BTW the best thing to happen to China AI is your immigration policy...”
Brad Feld
@bfeld
The U.S. leads in artificial intelligence, but for how long? www.technologyreview.com/s/609610/the-us-leads-in-artific…
-
Ian Goodfellow
@goodfellow_ian
#nips2017 At 4:45 PM I will speak about overcoming limited data with GANs in Grand Ballroom B www.iangoodfellow.com/slides/2017-12-09-label.pdf (This corrects an earlier tweet and the nips.cc page saying 4:15, the workshop site says 4:45 and is correct)
-
Ian Goodfellow
@goodfellow_ian
Sorry, two schedules got out of sync. It will be 4:45
-
Ian Goodfellow
@goodfellow_ian
Don't try this at home
Victoria Krakovna
@vkrakovna
@goodfellow_ian on the challenges of value alignment at #nips2017
-
Ian Goodfellow
@goodfellow_ian
A poetic outcome: @L_badikho found out about @black_in_ai because of people questioning the need for @black_in_ai on twitter
El Mahdi El Mhamdi
@L_badikho
Hello, thanks to all who participated in the push-back (and made some of us aware of @black_in_ai existence).
BiAI was a big success and…should we even mention the party?
great job @timnitGebru @red_abebe @negar_rz and everyone I didn't mention who made this possible!
#nips2017 -
Ian Goodfellow
@goodfellow_ian
I had a great time at @black_in_ai . Really felt like a historic turning point
Black in AI
@black_in_ai
Thank you to all who attended the workshop yesterday! And many thanks to fantastic keynote speakers, contributed talk presenters, poster presenters, & our sponsors! with @red_abebe, @BrownSarahM, Moustapha Cisse, @timnitGebru, @sanmikoyejo at #nips2017 #nips #ai #machinelearning
-
Ian Goodfellow reposted
Jack Clark
@jackclarkSF
Probably the scariest thing about #nips2017 is the undeniably massive effects that will come from scaling up existing AI components via incoming compute substrates from Graphcore/TPU2s, etc, and broad ignorance of this trend among policymakers. (I'm trying to change that!)
-
Ian Goodfellow
@goodfellow_ian
I had a great time visiting the CAIS++ student group at USC during NIPS. Many thanks for inviting me!
USC CAIS++
@CAISplusplus
A huge thank you to @goodfellow_ian of Google Brain for taking the time to stop by USC and chat with our club! Consider us inspired 🚀
-
Ian Goodfellow
@goodfellow_ian
#nips2017 I will speak about bridging the theory and practice of GANs in Hall A at 9:15
-
Ian Goodfellow reposted
Maithra Raghu
@maithra_raghu
Deep Learning: Bridging Theory and Practice happening tomorrow at #nips2017! Final Schedule: ludwigschmidt.github.io/nips17-dl-workshop-website/ @lschmidt3 @OriolVinyalsML @rsalakhu We have an exciting program with talks by Yoshua Bengio @goodfellow_ian Peter Bartlett Doina Precup Percy Liang Sham Kakade!
-
Ian Goodfellow
@goodfellow_ian
#nips2017 Which will we solve first?
-
Ian Goodfellow
@goodfellow_ian
GANs for personalized "predictive fashion": www.technologyreview.com/s/609469/this-ai-learns-your-fas…
-
Ian Goodfellow
@goodfellow_ian
GANs for generating images of how clothes will fit. Only two of these images are photos. qz.com/1090267/artificial-intelligence-can-now-show-you-h…
-
Ian Goodfellow
@goodfellow_ian
The first Black in AI event will take place at NIPS 2017: ai.stanford.edu/~tgebru/blackAI
-
Ian Goodfellow
@goodfellow_ian
Face swapping videos with CycleGAN: www.youtube.com/watch?v=Fea4kZq0oFQ
-
Ian Goodfellow
@goodfellow_ian
Adversarial examples that make stop signs misclassified 100% of the time, from multiple viewpoints: arxiv.org/pdf/1707.08945v1.pdf
-
Ian Goodfellow
@goodfellow_ian
Apple has launched an online machine learning journal, similar to Distill. First article is on GANs! machinelearning.apple.com/2017/07/07/GAN.html
-
Ian Goodfellow
@goodfellow_ian
NIPS 2017 adversarial examples challenge: www.kaggle.com/c/nips-2017-non-targeted-adversarial-attack Test your defenses against others' adversarial examples and vice versa!
-
Ian Goodfellow
@goodfellow_ian
New blog post with @NicolasPapernot on why verification methods for machine learning are so difficult: www.cleverhans.io/security/privacy/ml/2017/06/14/verifica…
-
Ian Goodfellow
@goodfellow_ian
CycleGAN learns to turn horses into zebras *without supervision*: arxiv.org/pdf/1703.10593.pdf











































































































































































































































