
MIT's Computer Science & Artificial Intelligence Laboratory (CSAIL). Media Inquiries: [email protected] Check out the latest CSAIL content ⬇️
Cambridge, MA
Joined October 2009
- Tweets6.8K
- Following20.4K
- Followers352K
- Likes2.8K
MIT CSAIL retweeted
AI is quickly improving at tough theory questions in math, physics, & economics.
In a new MIT study, frontier OpenAI models’ success rate rose from 3% to 35% in 16 months, while a successful attempt cost about $15. The new bottleneck may be verification: bit.ly/3UoRJee
MIT CSAIL retweeted
Want genAI to make you a functional, real-world object? Good luck — their 3D models are often physically flawed & hard to edit.
New "InstructMesh" tool makes it easy to generate, tweak, & 3D print things that actually work, such as decor & medical devices: bit.ly/4z5L5bu
MIT CSAIL retweeted
As humanoid robots become part of our everyday lives, they need to do more than just perform impressive motions. They should be safe, fun to interact with, and respond naturally to unexpected contact, without sacrificing performance.
Our latest work, Compliant GMT (C-GMT), brings whole-body compliant responses to general motion tracking.
🌐 More details: venki1899.github.io/c-gmt/
Work done at MIT Improbable AI Lab with @michl_wang, Jin, @pulkitology.
MIT CSAIL retweeted
"DREAM" is an AI model from MIT & Meta that can both understand images & generate them from text — then judge its own rough drafts to finish the best one.
The result: stronger vision, better image quality, & ~10% faster generation w/external re-rankers: bit.ly/4hBJing
MIT CSAIL retweeted
Left: MIT computer scientist Katie Bouman w/stacks of hard drives of black hole image data.
Right: MIT computer scientist Margaret Hamilton w/the code she wrote that helped put a man on the moon.
(image credit @floragraham)
#EHTblackhole #BlackHoleDay #BlackHole
MIT CSAIL retweeted
We’ve lost an icon. Margaret Hamilton, a computer science pioneer best known for her work at MIT leading development of the flight software for the Apollo missions to the moon, has died at 90. Read about her legacy via MIT News: news.mit.edu/2026/margaret-h…
MIT CSAIL retweeted
There’s lots of buzz around agentic harnesses for robots, particularly for long horizon tasks that require complex reasoning and memory. But what will it really take to turn a reasoning agent into a reactive, reliable and low-latency robot policy?
In our new paper, workspace models, we design a new memory architecture that acts as a latent harness for stronger reasoning models. Here’s why we think this might be the way forward:
(1/9)
MIT CSAIL retweeted
Can “chicken” make a base model reason? 🐔
Yes!
Why? With the right first tokens, base models can match RL in reasoning. We trace this to learned training data associations and use a simple data edit to make “chicken” elicit reasoning too.
blog: sophielwang.com/cues
MIT CSAIL retweeted
There is a new kind of robotics called "Vibe Robotics"
It's when you ask agents for fully functioning embodied agents (both hardware and software) that must solve a given task in physically grounded environments.
We build an evaluation to test how good agents (frontier models) are at "Vibe Robotics".
Website: artifactarena.ai/
MIT CSAIL retweeted
Students sometimes ask me if it still makes sense, in this accelerating age, to pursue a PhD in AI.
Perhaps counterintuitively, I think it's a great time to do so.
I wrote up some thoughts on this here: web.mit.edu/phillipi/www/wri…
MIT CSAIL retweeted
Did you know the original Claude — the chatbot’s namesake, according to Anthropic lore — was a renowned MIT professor? Claude Shannon SM ’40, PhD ’40, who taught at MIT from 1956 until his retirement in 1978, was known to many as the father of information theory and of modern digital communications.
MIT CSAIL retweeted
Today @Google, we published work led by @m_codreanu on AI in Science, drawing on 15 million Gemini interactions, 2,600 specialised AI models & a 600-scientist survey. What did we find about how AI is changing science? Read on...
MIT CSAIL retweeted
Excited to share our Cross-Species Single-cell Atlas of the Human Brain's Striatum, the brain region that integrates action planning, reward, motivation, decision-making, movement, learning, and reinforcement, implicated in schizophrenia, depression, addiction, and Huntington's Disease: authors.elsevier.com/c/1niW9…
Our map spans 109 samples, 69 donors, and 670,000 nuclei. We identify 31 neuronal subpopulations, including 9 types of medium spiny neurons, and rare D1 interface MSNs forming island-like structures. Patch-clamp recordings show outlier MSNs have smaller somata and stronger hyperpolarization-activated currents than their neighbors.
We show that the dorsolateral-ventromedial gradient shapes: receptor repertoires, disease gene expression, vulnerability.
Across 124 GWAS, we show genetic risk for neuropsychiatric, cognitive and addiction traits concentrates in MSNs and is depleted in glia and vascular cells. Schizophrenia, bipolar and ADHD risk genes are dorsally enriched, and so, unexpectedly, are opioid use disorder risk genes — even though opioid signaling itself is ventrally biased, suggesting heritable risk and acute drug action may engage different parts of the striatum.
Only a quarter of the human gradient genes are similarly zonated in mice. Several drug targets sit in different cells across species, including the mu-opioid receptor, the M5 muscarinic receptor, and the GLP-1 receptor, with important implications for preclinical screening. Profiling chronic antipsychotic exposure in mice, we find the transcriptional response strongly biased toward ventral MSNs.
Finally, we connect genotype to phenotype at single-cell resolution. Our new method, scMOSAIC, reads HTT CAG repeat length and gene expression from the same nucleus — and finds that repeat length varies along that same dorsolateral-ventromedial axis.
Huge congrats to Raleigh Linville @LinvilleRaleigh , Benjamin James @__ben_james__ , who carried this from start to finish, and to co-senior authors #MyriamHeiman and #DanaGabuzda, and Kiki Galani, Li-Lun Ho, Stu Fass, James Cameron, Brent Fitzwalter, Erica Engelberg-Cook, Sebastian Pineda, Dennis Dickson, Deborah Mash, Gustavo Turecki, Vanessa Wheeler, and Veronica Alvarez, across Massachusetts Institute of Technology, The Picower Institute for Learning and Memory, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), Broad Institute of MIT and Harvard, Dana-Farber Cancer Institute, The National Institutes of Health, Mass General Brigham, Mayo Clinic, McLean Hospital, McGill University, and beyond.
Huge thanks to the #SCORCH consortium and the brain banks in the US and Canada, and a huge thanks to the donors and their families. Every cell in this atlas came from someone who decided their brain should go on being useful. The data are public, and we hope other labs take them further than we can.
















