We published new research on how we serve post-trained Qwen3 235B models on NVIDIA GB200 NVL72 Blackwell racks.
GB200 is a major step up over Hopper for high-throughput inference on large MoE models, not just a training platform.
Ashish Vaswani
Ashish Vaswani is an Indian computer scientist and entrepreneur. He conducted research at Google Brain, co-founded Adept AI, and was co-founder and chief executive officer of Essential AI. As of June 2026, Vaswani, along with Essential AI, was acqui-hired by Nvidia. More on Wikipedia
-
Ashish Vaswani reposted
Perplexity
@perplexity_ai
-
Ashish Vaswani
@ashVaswani
Rnj-1 aces telco benchmarks. Congratulations to Farbod Tavakkoli, @GregoryDiamos , @tensorwave , @AMD , and the @essential_ai team!
Essential AI
@essential_ai
Rnj-1’s performance is especially good in correctness and abstention in its weight class, which are the two most important metrics for this work.
-
Ashish Vaswani
@ashVaswani
We’ve made a few updates based on feedback.
Essential AI
@essential_ai
[1/5]
We thank the community for their feedback on Rnj-1. We’d like to announce a few updates to Rnj-1-instruct based on what we heard:
- Resolving premature truncation of generations and improved instruction following.
- Instructions for 128k context length extrapolation.
- Updated evals, baselines, and model generations for reproducibility.
Details follow 🧵 -
Ashish Vaswani reposted
Adarsh Chaluvaraju
@_adarshc
Over a week ago, we at @essential_ai released RNJ-1, a world-class 8B model for code and math.
It’s now ranked #1 and #8 in text generation on Hugging Face.
📥 Instruct: 447k downloads — enthusiasts running the model locally, and providers enabling access to it.
📥 Base: 139k downloads — with businesses and researchers building on top of it. -
Ashish Vaswani reposted
Essential AI
@essential_ai
Rnj-1 is now live on @openrouter!
-
Ashish Vaswani
@ashVaswani
Rnj-1-Instruct is now the #1 trending text generation model on HF!
Ashish Vaswani
@ashVaswani
Tremendous effort by the @essential_ai team!
-
Ashish Vaswani reposted
ollama
@ollama
.@essential_ai's rnj-1 model is now on Ollama!
ollama run rnj-1
8B parameter, open-weight dense model trained from scratch. The model is optimized for code and STEM with capabilities on par with other state of the art open-weight models.
Let's go! 🚀🚀🚀 -
Ashish Vaswani
@ashVaswani
Cool demos!
Yash Jain
@_jainyash
The amount of bugs I have asked rnj-1 to fix for me are uncountable. So happy to share it with people!
-
Ashish Vaswani
@ashVaswani
It was a wonderful experience working with @vipulved and the @togethercompute team to bring Rnj-1 to the world.
Vipul Ved Prakash
@vipulved
The OSS coding model that @ashVaswani and team at @essential_ai have been working on over the last year is now out and available to use on @togethercompute APIs and playground.
It's particularly exciting that @essential_ai is doing fundamental work in pre-training and post training and publishing open weights!
www.together.ai/models/rnj-1-instruct
essential.ai/research/rnj-1 -
Ashish Vaswani
@ashVaswani
We are beyond thrilled to share our first flagship models, Rnj-1 base and instruct 8B parameter models. Rnj-1 is the culmination of 10 months of hard work by a phenomenal team, dedicated to advancing American SOTA OSS AI.
Lots of wins with Rnj-1.
1. SWE bench performance close to GPT 4o.
2. Tool use outperforming all comparable open source models.
3. Mathematical reasoning (AIME’25) nearly at par with GPT OSS MoE 20B.
….
Essential AI
@essential_ai
Today, we’re excited to introduce Rnj-1, @essential_ai's first open model; a world-class 8B base + instruct pair, built with scientific rigor, intentional design, and a belief that the advancement and equitable distribution of AI depend on building in the open.
We bring American open-source at par with the best in the world.Post not available here, view on X
Ashish Vaswani
@ashVaswani
Thanks to @Divyasmansingka for operational support.
Ashish Vaswani
@ashVaswani
Honored and grateful to work with an indefatigable team and brilliant leads @_saurabh, @pcmonk @kurtosis0 @Khoi_H_Nguyen @DivShivaprasad @peter_rushton.
Read more about our work on our blog and we will share a technical report soon.
www.essential.ai/research/rnj-1 -
Ashish Vaswani reposted
Athul Paul Jacob
@apjacob03
Today marks an important milestone.
I’m launching Percepta together with @htaneja, @hirshjain, @tmathew0309, Radha Jain, @marisbest2, @KonstDaskalakis and an incredible team, with the goal of bringing AI to the core industries that run our economy. For AI to deliver transformational impact, we need a strategy that orchestrates multiple disciplines. The most meaningful work of my career has been with strong interdisciplinary teams who take on ambitious challenges.
That’s the kind of team we’re building at Percepta.
Two pillars key pillars our work:
First, we deeply partner. Our researchers, product managers, and engineers work inside our partners’ operations - shadowing workflows, mapping decision points, building AI systems, and ultimately owning outcomes.
Second, we believe organizations revolve around decisions, and we’re reimagining decision-making to orchestrate AI agents and humans. This is our core research focus, blending foundation models, reinforcement learning, and optimization. The unique challenges of these domains won’t be solved by mimicking human data or finding a better prompt. Decision-making has many distinct elements, and our research team brings complementary skills to the challenge: collectively, we’ve built superhuman bots, deployed RL agents at scale, trained large language models to reason, proved fundamental results in learning theory, and released algorithms and tools used by the field every day.
We’ve already seen outcomes that change behavior: step-function productivity improvements in financial services, material enterprise value in care delivery organizations, and orders-of-magnitude cycle-time reductions in public-sector processes.
A bit over a year ago today, I joined @generalcatalyst - compelled by @htaneja vision to partner with him and help establish the early workings of Percepta. Since then, we’ve been quietly building a team of 25+.
We’re based in NYC and Boston and growing quickly. I’m grateful to the broader @generalcatalyst famiglia and especially @quentinclark, Jeannette zu Fürstenberg, @marcbhargava, David Fialkow, and @teresacarlson for being early partners in our journey.
We’re working against an incredibly ambitious mission. If this excites you, let’s chat!
Learn more:
www.percepta.ai
www.generalcatalyst.com/stories/unveiling-percepta -
Ashish Vaswani
@ashVaswani
Proud of @RitvikKapila and team’s collaboration with DAFT. It really simplified our web-scale data processing.
Ritvik Kapila
@ritvikkapila
1/We, at @essential_ai recently released Essential-Web v1.0 -- a 24T token dataset labelled by a 12 category taxonomy, produced by EAI-Distill-0.5b, a fine-tuned 0.5b-parameter model. With simple SQL filters, we obtain competitive datasets in math, web code, STEM and medical.
-
Ashish Vaswani
@ashVaswani
At Essential AI, we strive to do Foundational work in the open. Open Science ensures the sustained progress of AI.
Essential AI
@essential_ai
[1/3]
The White House AI Action Plan makes a clear case:
“Open-source and open-weight models could become global standards in business, science, and academic research worldwide; they carry immense geostrategic significance.”
This aligns deeply with our mission and values at Essential AI. -
Ashish Vaswani reposted
clem 🤗
@ClementDelangue
We just released the best 3B model, 100% open-source, open dataset, architecture details, exact data mixtures and full training recipe including pre-training, mid-training, post-training, and synthetic data generation for everyone to train their own.
Let's go open-source AI!
Loubna Ben Allal
@LoubnaBenAllal1
Introducing SmolLM3: a strong, smol reasoner!
> SoTA 3B model
> dual mode reasoning (think/no_think)
> long context, up to 128k
> multilingual: en, fr, es, de, it, pt
> fully open source (data, code, recipes)
huggingface.co/blog/smollm3 -
Ashish Vaswani reposted
Aurko Roy
@aurko79
Excited to share what I worked on during my time at Meta.
- We introduce a Triton-accelerated Transformer with *2-simplicial attention*—a tri-linear generalization of dot-product attention
- We show how to adapt RoPE to tri-linear forms
- We show 2-simplicial attention scales better under token constraints than dot product attention
It was fun collaborating with amazing folks including @dvsaisurya @_arohan_ and others -
Ashish Vaswani reposted
Daniel Campos
@spacemanidol
I am trying out this Thought-Boi Thing. Give it a read.
The Hidden Cost of Augmentation: Every Tool You Use Changes You. open.substack.com/pub/spacemanidol/p/the-hidden-cost-of-a… -
Ashish Vaswani
@ashVaswani
We didn't see clear wins on grokking with Muon. Curious if others have observed similar behaviors.
Essential AI
@essential_ai
[1/5]
We have a quick update to share, which contradicts our hypothesis regarding the abilities of Muon and Adam vis-a-vis Grokking.
Ashish Vaswani
@ashVaswani
This was a fun collaboration with the superb @ssingla17 and @ishaankshah, who led the effort.
-
Ashish Vaswani
@ashVaswani
💪
Ritvik Kapila
@ritvikkapila
#1 trending on @huggingface letsgoooo!
@essential_ai 🥇 -
Ashish Vaswani
@ashVaswani
Check out our latest research on data. We're releasing 24T tokens of richly labelled web data. We found it very useful for our internal data curation efforts.
Excited to see what you build using Essential-Web v1.0!
Essential AI
@essential_ai
[1/5]
🚀 Meet Essential-Web v1.0, a 24-trillion-token pre-training dataset with rich metadata built to effortlessly curate high-performing datasets across domains and use cases!
Ashish Vaswani
@ashVaswani
Terrific work by @AndrewHojel , Michael Pust, Tim Romanski, and the Data Team at @essential_ai !
-
Ashish Vaswani reposted
Andrew Hojel
@AndrewHojel
Check out what the Data Team at @essential_ai has been cooking! It's been a blast preparing this dataset and super excited to see what people use it for. Shoot me a DM with any questions or cool use cases.
Essential AI
@essential_ai
[1/5]
🚀 Meet Essential-Web v1.0, a 24-trillion-token pre-training dataset with rich metadata built to effortlessly curate high-performing datasets across domains and use cases! -
Ashish Vaswani reposted
Yash Vanjani
@YashVanjani
Super excited to share how we achieved significant performance gains in Muon optimizer for Pre-training LLMs at @essential_ai!
This was a great team effort — special thanks to @pcmonk , whose core ideas were instrumental in driving this improvement, and to @ashVaswani @ishaankshah @karlstratos for their close collaboration and support throughout.
Essential AI
@essential_ai
Muon is a serious competitor to AdamW, but it's tricky to scale up.
Our infra team has made fundamental advancements in parallelizing Muon on large scale distributed clusters. We're extremely happy with the result and it's now a part of our pretraining pipeline.
🔗Check out our new blog post on this:
www.essential.ai/blog/infra
www.essential.ai
@pcmonk @YashVanjani @ishaankshah @karlstratos @ashVaswani -
Ashish Vaswani
@ashVaswani
Terrific work speeding up Muon led by @pcmonk, in collaboration with @YashVanjani , @ishaankshah , and @karlstratos . We’re exited about the benefits we’re seeing with Muon.
Essential AI
@essential_ai
Muon is a serious competitor to AdamW, but it's tricky to scale up.
Our infra team has made fundamental advancements in parallelizing Muon on large scale distributed clusters. We're extremely happy with the result and it's now a part of our pretraining pipeline.
🔗Check out our new blog post on this:
www.essential.ai/blog/infra
www.essential.ai
@pcmonk @YashVanjani @ishaankshah @karlstratos @ashVaswani -
Ashish Vaswani reposted
clem 🤗
@ClementDelangue
Cool paper from @essential_ai @ashVaswani huggingface.co/papers/2505.02222
-
Ashish Vaswani
@ashVaswani
Please check out our thorough study on the advantages of Muon. Second-order optimization is a promising path to more efficient LLM pretraining.
Essential AI
@essential_ai
[1/5]
Muon has recently emerged as a promising second-order optimizer for LLMs. Prior work (e.g. Moonshot) showed that Muon scales. Our in-depth study addresses the practicality of Muon vs AdamW and demonstrates that Muon expands the Pareto frontier over AdamW on the compute-time tradeoff. -
Ashish Vaswani reposted
Essential AI
@essential_ai
🗞️ We just launched our new landing page and dropped a fresh blog post on how LLMs learn to reflect and revise their thinking: In order to advance reasoning, it's vital to measure and understand its constituents, such as reflection.
More to come - www.essential.ai -
Ashish Vaswani
@ashVaswani
Reinforcement learning has shown success in eliciting reflection from LLMs, but what if this capability actually manifests earlier in pre-training? We investigated this question and our results are surprising 👇
[1/4]Post not available here, view on X
Ashish Vaswani
@ashVaswani
A big lift in performance comes from a simple cue: “Wait,”. This trigger prompts explicit reflection-especially as pre-training progresses-and matches the performance of models that are explicitly told there’s an error. It’s a causal signal for reflection improving accuracy.
[3/4]
Ashish Vaswani
@ashVaswani
To learn more, read our complete paper. We will be sharing additional results with the community in the coming weeks because we believe open science will accelerate the development of frontier capabilities.
Paper Link: arxiv.org/abs/2504.04022
[4/4] -
Ashish Vaswani
@ashVaswani
We are hiring our first product engineers. Come join an incredible team advancing Human AI interaction!
Please check our jobs page for all the open positions
jobs.ashbyhq.com/essentialai -
Ashish Vaswani reposted
Delip Rao
@deliprao
If this were a science paper, you would expect a country that picks its science workforce at random as a “weak baseline” and a leading nation like the US to actively experiment towards state-of-the-art, or at least beat the baseline.
Not providing a guaranteed path for accomplished scientists like @rdesh26, who graduated from one of the top labs in the country, @jhuclsp, to contribute to our progress and security directly is an avoidable tragedy.
I implore the @WhiteHouse / @USAGov OSTP, @StateDept, and @STASatState to look into the ways we are leaching out the best talent in the country, starting with this lottery system for the nation’s most talented.
Many have made this request before to past administrations to no avail, but acting on this now would leave a legacy that will be the envy of future administrations. I hope you do. If you have doubts, ask other best scientists nationwide, including @theNASciences. -
Ashish Vaswani
@ashVaswani
I'm thrilled to announce our company, @essential_ai . We believe that breakthroughs in AI will unlock the most profound tools for thought, advancing humanity's collective knowledge and capability.
essential.aiPost not available here, view on X
Ashish Vaswani
@ashVaswani
We are thankful for the support from our angels: @amasad, @altcap, @w_conviction, @eladgil, @fdesouza , David H. Patraeus, @GSapoznik, @jwmontgomery, and Mei Zuo.
Ashish Vaswani
@ashVaswani
Follow along at @essential_ai or visit us at essential.ai to stay updated with our progress.
-
Ashish Vaswani reposted
Tri Dao
@tri_dao
The tech report has all the info: tridao.me/publications/flash2/flash2.pdf
More details in blogposts:
crfm.stanford.edu/2023/07/17/flash2.html
together.ai/blog/tri-dao-flash-attention
princeton-nlp.github.io/flash-atttention-2/
FlashAttention-2 is available in the open source: github.com/Dao-AILab/flash-attention
2/ -
Ashish Vaswani reposted
-
Ashish Vaswani reposted
Perplexity
@perplexity_ai
Introducing Bird SQL, a Twitter search interface that is powered by Perplexity’s structured search engine. It uses OpenAI Codex to translate natural language into SQL, giving everyone the ability to navigate large datasets like Twitter.
www.perplexity.ai/sql -
Ashish Vaswani reposted
Perplexity
@perplexity_ai
Announcing Perplexity Ask, a new search interface that uses OpenAI GPT 3.5 and Microsoft Bing to directly answer any question you ask.
www.perplexity.ai
discord.com/invite/kWJZsxPDuX -
Ashish Vaswani reposted
Nathan Benaich
@nathanbenaich
🪩The @stateofai 2022 is live!🪩
In its 5th year, the #stateofai report condenses what you *need* to know in AI research, industry, safety, and politics. This open-access report is our contribution to the AI ecosystem.
Here's my director's cut 🧵:
www.stateof.ai -
Ashish Vaswani reposted
Neeva
@Neeva
10/ We are releasing our model (huggingface.co/neeva/query2query) and golden set used for eval (huggingface.co/datasets/neeva/query2query_evaluation) on @huggingface.
Take a look at our latest blog post for more information 👁️ ⤵️ neeva.com/blog/state-of-the-art-query2query-similarity -
Ashish Vaswani
@ashVaswani
We've taught Transformers to take actions in the digital world!
-
Ashish Vaswani reposted
Andrew Carroll
@acarroll_ATG
DeepConsensus is published! You can use this link if you need - rdcu.be/cUMIx
Excited to see transformers used for DNA error correction, and scientists getting more out of their sequencing instruments.
Nature Biotechnology
@NatureBiotech
DeepConsensus improves the accuracy of sequences with a gap-aware sequence transformer go.nature.com/3q8kiK0
-
Ashish Vaswani reposted
-
Ashish Vaswani reposted
Saam Motamedi
@saammotamedi
Adept is moving fast in building usable general intelligence.
In the last 3 months, it has learned how to query databases, fetch public data, and analytics and visualization.
It will soon learn all the software tools we use as it transforms how we work.
Check out the video👇 -
Ashish Vaswani
@ashVaswani
We’re building the future of useful and intelligent machines. Our early neural networks can make plots, query databases and fetch internet data! If you’re excited to work on fundamental research like learning from human feedback and building sample efficient models, please apply!
-
Ashish Vaswani
@ashVaswani
After 5+ wonderful years in Google Brain, working at the forefront of ML alongside inspiring colleagues, I'm excited to share my new adventure. We started Adept with the mission to build the future of human-computer collaboration. www.adept.ai/about-us.
Ashish Vaswani
@ashVaswani
Please reach out to me if you're interested in working with us!
-
Ashish Vaswani reposted
Mostafa Dehghani
@m__dehghani
With @YiTayML, @anuragarnab, @giffmana, and @ashVaswani, we wrote up a paper on "the efficiency misnomer": arxiv.org/abs/2110.12894
TL;DR:
"No single cost indicator is sufficient for making an absolute conclusion when comparing the efficiency of different models". -
Ashish Vaswani reposted
tsvetshop
@tsvetshop
"Simple and Efficient ways to Improve REALM"
by @vidhisha_b, @ashVaswani, Yulia Tsvetkov and @nikiparmar09
in Workshop on Machine Reading for Question Answering #emnlp2021
Preprint: arxiv.org/abs/2104.08710 -
Ashish Vaswani reposted
Gunjan Baid
@gunjan_baid
Very excited to present our work on DeepConsensus, a transformer model for PacBio HiFi sequencing. The model operates on a MSA of subreads + draft consensus and was trained using an alignment-based loss. Collaboration with many amazing folks @Google and @PacBio. Congrats, team!
Andrew Carroll
@acarroll_ATG
We’re excited to release DeepConsensus, a transformer approach for PacBio HiFi consensus generation. DeepConsensus reduces HiFi read errors by 42%, improving assembly NG50, base accuracy, as well as variant calling. Led by @gunjan_baid and @daniel_e_cook
www.biorxiv.org/content/10.1101/2021.08.31.458403v1 -
Ashish Vaswani reposted
niccolò zanichelli
@nc_znc
Sharing my implementation of @AravSrinivas et al's Bottleneck Transformers for Visual Recognition (arxiv.org/abs/2101.11605) in JAX/Flax.
Check it out here: github.com/NZ99/bottleneck-transformer-flax -
Ashish Vaswani
@ashVaswani
(1/5) In our recent CVPR paper, we develop a new family of parameter-efficient local self-attention models, HaloNets, that outperform EfficientNet in the parameter-accuracy tradeoff on ImageNet. arxiv.org/pdf/2103.12731.pdf.
Post not available here, view on X
Ashish Vaswani
@ashVaswani
(4/5) When applied to detection and instance segmentation, our local self-attention improves on top of strong convolutional baselines. Interestingly, local self-attention with 14x14 receptive fields performs nearly as well as 35x35.
Ashish Vaswani
@ashVaswani
(5/5) Joint work with Prajit Ramachandran, @AravSrinivas , @nikiparmar09 , @BlakeHechtman , and Jonathon Shlens.
-
Ashish Vaswani reposted
rohan anil
@_arohan_
Code for Distributed Shampoo: a scalable second order optimization method
github.com/google-research/google-research/tree/master/sc… 💥
Joint work w @GuptaVineetG
State of the art on MLPerf ResNet-50 training to reach 75.9% accuracy at 32,768 batch size
Trains in 1729 steps (not a typo), 284 secs on TPUs. -
Ashish Vaswani reposted
sageev
@osageev
This is so fun: someone recorded the original #transformers film theme song into our #speech2music demo (which uses a music transformer model) and it is pretty cool to listen to :) (recommend using the MediumV1 sparsifier option):
panther.research.cs.dal.ca/index.html
@ashVaswani @huangcza
sageev
@osageev
In this #speech2music #neurips2020 #interactive #demo, you can record your own #speech and have it run through a specialized #music transformer model to #generate new musical material!
Interactive site: jasondeon.github.io/musicalSpeech/
Demo, Samples & Workshop links to follow in thread -
Ashish Vaswani reposted
Nathan Benaich
@nathanbenaich
💥We're live with @stateofai 2020💥
For the 3rd year, @soundboy and I spent 2 months analyzing the most interesting developments in AI.
We aim to trigger an informed conversation about the #stateofai and its implication for the future. Thread 👇
www.stateof.ai -
Ashish Vaswani reposted
Nathan Benaich
@nathanbenaich
I wrote up an opinion for the @FT on why AI has yet to make a big impact on our fight against the coronavirus.
I’d love to hear your opinions, please drop them in the comments below.
cc @airstreet @stateofai @raais @londonai
Financial Times
@FT
Why AI has disappointed on Covid www.ft.com/content/0aafc2de-f46d-4646-acfd-4ed7a7f6feaa
-
Ashish Vaswani reposted
hardmaru
@hardmaru
Efficient Content-Based Sparse Attention with Routing Transformers
Self-attention with a sparse routing can reduce overall complexity of attention from O(n² d) to O(n√n d), while getting great results on Wikitext-103, enwik8, and ImageNet-64 generation.
arxiv.org/abs/2003.05997 -
Ashish Vaswani reposted
OpenAI
@OpenAI
Congrats to our very own @woj_zaremba on being named one of this year's MIT Tech Review Innovators Under 35 for his achievements in Robotics! www.technologyreview.com/lists/innovators-under-35/2019/p…
-
Ashish Vaswani reposted
Arvind Neelakantan
@arvind_io
We describe a simple technique to parallelize Scheduled Sampling across time that allows us to apply Scheduled Sampling for problems that involve generating very long sequences. We get better sample quality and train almost as fast as teacher-forcing.
arxiv.org/abs/1906.04331 -
Ashish Vaswani
@ashVaswani
Pure content based interactions are competitive for vision models. Lot's of exciting work to be done in this research area.
niki parmar
@nikiparmar09
New Paper:
Stand-Alone Self-Attention in Vision Models
arxiv.org/abs/1906.05909
Can attention work as a stand-alone primitive for vision models?
We develop a pure self-attention model by replacing the spatial convolutions in a ResNet by a simple, local self-attention layer.
Ashish Vaswani
@ashVaswani
Our self-attention model outperforms the Resnet baseline on ImageNet classification and matches RetinaNet on object detection with fewer FLOPS and parameters.
-
Ashish Vaswani reposted
rapha
@rapha_gl
2nd @GoogleAI residency paper now on arxiv!
"Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation"
👉🏽 arxiv.org/abs/1906.02611
Patch Gaussian gets SOTA on Common Corruptions & more!
Thread for deets 👇🏽 -
Ashish Vaswani reposted
Jason Baldridge
@jasonbaldridge
Side note: the VLN stuff will get really fun as it gets more and more interactive, and even involves games. The language will become more urgent and interesting, and having dialogue with real time constraints will make language matter more than with static instruction following.
-
Ashish Vaswani reposted
Jason Baldridge
@jasonbaldridge
New ACL paper on the Room-to-Room (R2R) dataset by @vihaniaj, @gabriel_ilharco, @alex_y_ku, @ashVaswani, Eugene Ie and me, "Stay on the Path: Instruction Fidelity in Vision-and-Language Navigation." The tag line is "putting the L back in VLN". (thread)
arxiv.org/abs/1905.12255 -
Ashish Vaswani reposted
Quoc Le
@quocleix
Exciting new work on replacing convolutions with self-attention for vision. Our paper shows that full attention is good, but loses a few percents in accuracy. And a middle ground that combines convolutions and self-attention is better. Link: arxiv.org/abs/1904.09925
-
Ashish Vaswani reposted
Manaal Faruqui
@manaalfar
New York City ❤️
StanceGrounded
@_SJPeace_
Jews created a path holding signs of solidarity outside a mosque entryway in New York so when Muslims walked through to enter the mosque, they see the support.
THIS IS HUMANITY
Retweet ❤️ -
Ashish Vaswani reposted
Jeff Dean
@JeffDean
I was lucky enough to meet Karen Sparck Jones early in my career when I was a researcher working at Digital Equipment Corporation's Western Research Lab (DEC WRL). I & several other researchers had started to get interested in information retrieval (cont) www.nytimes.com/2019/01/02/obituaries/karen-sparck-jones-…
-
Ashish Vaswani reposted
Ilya Sutskever
@ilyasut
The thing I like most is just how simple the idea is. And that unsupervised learning is really truly starting to work.
OpenAI
@OpenAI
We've trained an unsupervised language model that can generate coherent paragraphs and perform rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training: blog.openai.com/better-language-models/
-
Ashish Vaswani reposted
Jeff Dean
@JeffDean
Very happy to see my colleague Geoffrey Hinton has been named a Companion of the Order of Canada. A well-deserved honor! Congrats, Geoff!
Google Canada
@googlecanada
🎉🇨🇦 Congratulations to the godfather of Deep Learning, AI pioneer and @Google Engineering Fellow Geoffrey Hinton on being named a Companion of the #orderofcanada. @GoogleAI @UofT
-
Ashish Vaswani reposted
Nal
@nalkalc
More compute, better architectures and novel robust orderings have fueled tremendous progress for AR image models. SPNs are our latest installment. Looking forward to samples from 2020!
paper: arxiv.org/abs/1812.01608
open reviews (scores 9,10,7): openreview.net/challenge?redirect=%2Fforum%3Fid%3DHylzTiC… -
Ashish Vaswani reposted
Andrej Karpathy
@karpathy
Discovering (paradoxically late in life) that I get more out of textbooks than books and that I don’t have to stop buying them just because I’m out of school. Good reading list pointers: www.lesswrong.com/posts/xg3hXCYQPJkwHyik2/the-best-textbo…
-
Ashish Vaswani reposted
Ilya Sutskever
@ilyasut
bot: 1; humans: 0
-
Ashish Vaswani reposted
Ilya Sutskever
@ilyasut
www.nytimes.com/interactive/2018/07/30/technology/robot-h… -- an NYT article about Dactyl, our block-rotating robot hand
-
Ashish Vaswani reposted
Kyle Kastner
@kastnerkyle
Excited about Anna Huang (et. al. w/ Magenta)'s work with a modified Transformer for generating piano roll-likes, a) works stunningly well, even on small corpora like Bach, and b) is published in workshop form! PAPER: docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbW… && AUDIO: relative-attention-music-nips-2018.s3-website-us-west-1.a…
-
Ashish Vaswani reposted
James Owers-Bardsley
@jamesowers
Holy. Shit. Expressive, interesting, and lengthy piano music generations. I genuinely got goosebumps listening to these. As far as I'm concerned, this is a huge step forward for creative ML. Cheers Anna Huang. Paper: docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbW… #icml2018
-
Ashish Vaswani reposted
niki parmar
@nikiparmar09
Come by our talk and poster today to hear more about Image Generation using Self Attention. Paper: arxiv.org/pdf/1802.05751.pdf
#icml2018
Dustin Tran
@dustinvtran
Check out the Image Transformer by @nikiparmar09 @ashVaswani others and me. Talk at 3:20p @ Victoria (Deep Learning). Visit our poster at 6:15-9:00p @ Hall B #217! proceedings.mlr.press/v80/parmar18a.html
-
Ashish Vaswani reposted
Durk Kingma
@dpkingma
Check out blog.openai.com/glow/, my work with @prafdhar on improving flow-based generative models with invertible 1x1 convolutions. youtu.be/exJZOC3ZceA
This media couldn't be displayed. See it on the original post. -
Ashish Vaswani reposted
Nal
@nalkalc
Exciting news: @Google Brain expands to Amsterdam 🚲🚣🌳 Looking forward to working on core AI challenges with @TimSalimans, Lasse Espeholt and researchers across Google and beyond!
For RE and RS roles, reach out or apply here: www.google.com/about/careers/applications/jobs/results
Great times ahead!
Jeff Dean
@JeffDean
Welcome to Google, @TimSalimans!
This post is unavailable. See it on X. -
Ashish Vaswani reposted
hardmaru
@hardmaru
Btw, I'm still trying to hire research scientists for Google Brain Tokyo. I'm really trying hard to hire a diverse team (in Japan, of all places!) since I believe strongly that is the right thing to do. If you know anyone outstanding, please contact me!
hardmaru
@hardmaru
Excited about expanding the Google Brain team in Tokyo. If you are excited to work on fresh and interesting AI research, please consider applying! www.google.com/about/careers/applications/jobs/results
-
Ashish Vaswani reposted
Google AI
@GoogleAI
Designing robots that can observe their surroundings and decide the best course of action, while reacting to unexpected outcomes, is exceptionally difficult. Visit our blog to see what researchers from Google and X are doing to tackle this challenge. research.google/blog/scalable-deep-reinforcement-learning…
-
Ashish Vaswani reposted
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

































