Stanford NeuroAI Lab (@NeuroAILab) profile picture

Neuroscience, artificial intelligence, and psychology research at Stanford University (PI: Dan Yamins)

Stanford, CA
Joined November 2017
Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)Photo by Stanford NeuroAI Lab (@NeuroAILab)
Stanford NeuroAI Lab retweeted
On the heels of GRH being proven true by AI, our group has had a bunch of work on both the theory of AI & its safety, so I've written a two-part @mldcmu blogpost series summarizing it: Part I (or why robots may inevitably become sentient via our Capable Agents & Contravariance theories): blog.ml.cmu.edu/2026/11/19/t… Part II (working towards a science of *provable* safety, using capable agents as a case study to go from barriers to guarantees): blog.ml.cmu.edu/2026/11/19/t…
12
29
1,250
Stanford NeuroAI Lab retweeted
had fun helping out with the PSIBorg demo at CoRL, led by @AtlasKazemian , @KlemenKotar , @wenlong_huang and @ieva_bagd ! 🤖 PSIBorg brings next-token prediction to robotics. A single model can be used for various robotics capabilities (policy, simulation, progress tracking, etc.) simply by changing how we prompt it.
Last week at CoRL, we ran a two-day live demo of our new robot model! We brought the robot and a bunch of objects, people in the audience brought some of their own and helped decide what the robot should do. Here is how it went:
3
14
510
Stanford NeuroAI Lab retweeted
Check out how our lab's demo at CoRL went! PSIBorg is our robotics foundation model that we trained from scratch
Last week at CoRL, we ran a two-day live demo of our new robot model! We brought the robot and a bunch of objects, people in the audience brought some of their own and helped decide what the robot should do. Here is how it went:
2
8
387
Check out how our lab's demo at CoRL went! PSIBorg is our robotics foundation model that we trained from scratch
Last week at CoRL, we ran a two-day live demo of our new robot model! We brought the robot and a bunch of objects, people in the audience brought some of their own and helped decide what the robot should do. Here is how it went:
2
8
387
Stanford NeuroAI Lab retweeted
Last week at CoRL, we ran a two-day live demo of our new robot model! We brought the robot and a bunch of objects, people in the audience brought some of their own and helped decide what the robot should do. Here is how it went:
3
7
4
37
2,271
Stanford NeuroAI Lab retweeted
This is a fun package by @lynde_wolf from our lab that creates figures in the visual style of old school cognitive psychology papers. lynde-wolf.github.io/cogra/i…
8
40
4
260
10,999
Stanford NeuroAI Lab retweeted
Thrilled to work with @MSKCancerCenter to advance our world models of patient biology! We @NOETIK_ai have seen that scale and diversity of high-quality, multimodal data is a principal driver of better world models -- models that predict which patients will respond to which therapeutics. This collaboration is a major step on our path to bringing every patient in distribution, so that all patients and all clinical trials can benefit from model-driven insight into disease biology.
Noetik is expanding the scale and diversity of data behind our cancer biology world models. We’re excited to announce a collaboration with @MSKCancerCenter, providing access to proprietary data from tens of thousands of patients, collected during routine clinical care. By enabling a more diverse and comprehensive view of cancer development, the new data will help us train and advance our models’ ability to predict therapeutic response for patient impact. Read more here: businesswire.com/news/home/2…
1
2
20
689
Stanford NeuroAI Lab retweeted
Doing a PhD and interested in World Models? Come work with me and the Meta FAIR World Model team for a year. Exciting research, hundreds of GPUs and free food are awaiting you.
8
19
1
638
36,356
Stanford NeuroAI Lab retweeted
Is vision better explained by generative or discriminative learning ? 🧠👁️ Our #NeurIPS2026 paper says: neither! With Joint Energy-based Models (JEMs), we trained one network across the generative↔discriminative continuum. Human alignment peaks in between. 🧵
3
18
2
106
5,068
BCS is hiring! Search now underway for tenure-track assistant professor across all areas of brain and cognitive science, potentially including computation, cognitive science, or experimental neuroscience. Deadline Dec. 1, 2026. academicjobsonline.org/ajo/j…
1
6
1
32
2,690
Stanford NeuroAI Lab retweeted
I'm hiring a Research Scientist at @StanfordAILab ! Lead research, write and manage grants, teach, and mentor PhD students in my lab. A real growth opportunity, working directly with me. 3-year role, in person, $125K–$225K base. Apply: careersearch.stanford.edu/hc…
11
92
3
639
82,661
Stanford NeuroAI Lab retweeted
Through my wonderful time at NVIDIA and Stanford, one question has kept me up at night: Should the future of AI be monolithic or plural? To answer this question, we’ve raised $110M to get started, backed by @MenloVentures, @MadronaVentures, @NVIDIA, @NEA, @KindredVentures,@Wing_VC, @aixventureshq, @Principal_VC, @Cisco, @Snowflake, among others. Today, Carlos (@guestrin) and I are stoked to launch Noeri, a neolab to push the frontier of plural intelligence. Noeri means “the plural of intelligence” or many intelligences. The future of AI we root for is plural: a healthy ecosystem of many intelligences, in all sizes, strengths, and quirks. Discovery, in particular, has always been plural: many curious minds, each devoted to their own question, often on roads less traveled. Our path there is self-improving AI that: -> learns continually on the job, closing the gap between what AI can do off the shelf and what the dynamic world demands -> learns to use other AIs as tools and make new tools to overcome its own limitations -> picks the right-sized model for your problem, at a fraction of the frontier cost Then the power of intelligence, from the craftsmanship of making AI to ownership, can be in the hands of all. @noeri_ai mission reflects who we are as the founding team. My career has been anything but linear, literally continual learning on the job. Carlos has devoted his career to building tools that empower people through AI. We are all a bit contrarian, a bit adventurous, and above all, big believers in pluralism. Join us to build this future together: noeri.ai/
201
173
109
1,676
947,286
Stanford NeuroAI Lab retweeted
Replying to @dwarkesh_sp
Really enjoyed this, and totally agree that sample efficiency is one of the biggest gaps between AIs and humans. One framing I've found helpful is to split it into two parts. The first is learning broadly useful knowledge from limited experience, which "data to reach a given loss" captures pretty well. The second is using that knowledge on new tasks without any task-specific training. Kids seem to manage both, and current models only get partway on each. I wrote about this in the first post of a short series, in case it's of interest: khaiaw.substack.com/p/data-e… The rest of the series is about our recent work on it. We trained a Zero-shot World Model (ZWM) on just one child's experience, and it gets pretty close to state-of-the-art on a range of tasks, zero-shot. The last post touches on something close to your argument: evolution as a hint that the priors behind data-efficient learning could themselves be found by search. The nanoGPT speedrun feels like an early, human-run version of that.
2
7
680
Stanford NeuroAI Lab retweeted
It is not easy to characterize the features represented by human language cortex. This work is a step toward doing so. Using small, interpretable feature sets, we explain language-network responses and show a shared feature basis across regions, with variation across individuals.
Excited to share updates to our recent preprint, where we use small sets of SAE features to characterize language responses in the brain. In a new analysis, we use the SAE feature tunings discovered by our method to search for that exact tuning profile more broadly across the brain. Without using any anatomical information in voxel selection, we show that voxels with the same tuning profile tend to cluster in the same anatomical area (more below). Reposting the whole thread below because my old twitter account was hacked 😬
4
35
1,813
Stanford NeuroAI Lab retweeted
I’m starting a weekly blog where I (metaphorically) cut open the putative black box architecture of the transformer! 🤖 Part 1: Are transformers that different from convolutional networks? 🤔 neuroailab.github.io/transfo… #Neuroscience #NeuroAI #Vision
1
4
27
1,004
Stanford NeuroAI Lab retweeted
LLMs do single-objective, next-token prediction. An LLM is a single model that does many tasks that can be expressed in text, steered by prompts, and scales predictably. The same lesson may apply to robotics. One foundation model does next-token prediction on policy, simulation, and progress tracking - just different prompts through one set of weights.
Robot Capabilities as Probabilistic Inference over Next Token Prediction We are doing a demo at CoRL debuting PsiBorg: an 8B universally promptable robot model for open-world manipulation! One model works simultaneously as a policy, a simulator, and a progress tracker. It is trained from scratch, developed fully in-house at #Stanford, not a VLA or WAM, and (tries to) work with new objects and new goal images you throw at it! Come try it at #CoRL2026! 🕛 Mon Nov 9 · 12–6:30 PM CST 🕙 Tue Nov 10 · 10 AM–6:30 PM CST 📍 Demo #2 · Lone Star (3rd floor), JW Marriott Austin 🌐 neuroailab.github.io/psi-web…
1
4
155
Stanford NeuroAI Lab retweeted
Congratulations to BAIR Researchers from Roberto Horowitz, @pabbeel and @KoushilSreenath labs whose paper "SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model," was awarded the Best Conference Paper at @corl_conf yesterday in Austin! #CoRL2026 corl.org/program/awards
SuperMimic has won the CORL’s best paper award. Thanks for everyone’s support!
2
8
135
12,521
Stanford NeuroAI Lab retweeted
Atlas is now available to everyone!! I’m very excited to see what users can build with our newest world model! 🤩🌎
We’re opening up access to Atlas to everyone today. We’ve built apps and API endpoints to help you use Atlas any way you can imagine: bring your ideas for building a 3D world and come away with keyframes, video clips, or 3D Gaussian splats.
108
307
51
2,715
180,630
Stanford NeuroAI Lab retweeted
Really honored to have our ROGUE benchmark showing the *incorrigibility* of frontier agents a few months *before* the now-famous HuggingFace incident, featured in the NYT! (and interviewed alongside the father of corrigibility, @So8res, no less!)
1
5
2
26
3,401
Stanford NeuroAI Lab retweeted
In neuroscience, a canonical microcircuit is thought to subserve all cortical computation. In AI, neural networks universally subserve intelligent behavior. Is there convergence between computational motifs in natural and artificial intelligence? Work with @dyamins 🧠🤖
4
36
7
178
21,200