The secret life of an LM is defined by its internal data types. Inner layers transport abstractions that are more robust than words, like concepts, functions, or pointers. In new work yesterday, @arnabsensharma.bsky.social et al identify a data type for *predicates*. bsky.app/profile/arn...

What does an LLM do when it translates from Italian "amore" to Spanish "amor" or French "amour"? That's easy! (you might think) Because surely it knows: amore, amor, amour are all based on the same Latin word. It can just drop the "e", or add a "u".

Looking forward to #COLM2025 tomorrow. DM me if you'll also be there and want to meet to chat.

There are a lot of interesting details that surface when you use SAEs to understand and control diffusion image synthesis models. Learn more in @wendlerc.bsky.social's talk.

On the Good Fight podcast w substack.com/@yaschamounk I give a quick but careful primer on how modern AI works. I also chat about our responsibility as machine learning scientists, and what we need to fix to get AI right. Take a listen and reshare - www.persuasion.community/p/david-bau

I love the 'opinionated' approach taken by Aaron + team in this survey. It captures the ongoing work around the central casual puzzles we face in mechanistic interpretability.

Who is going to be at #COLM2025? I want to draw your attention to a COLM paper by my student @sfeucht.bsky.social that has totally changed the way I think and teach about LLM representations. The work is worth knowing. And you can meet Sheridan at COLM, Oct 7! bsky.app/profile/sfe...

Announcing a broad expansion of the National Deep Inference Fabric. This could be relevant to your research...

The NDIF youtube talk series continues... Don't miss the fascinating talks on by Xu Pan and Josh Engels, on the NDIF youtube channel. www.youtube.com/channel/UCaQ...

In the wake of the Jimmy Kimmel firing: Do not underestimate the power of the truth. The truth is our superpower. davidbau.com/archives/202...

Monday: Trump tries to fire Fed Governor Lisa Cook (first time in 111 years). Thursday: CDC chief dismissed, four top scientists resign. Discredit, dismiss, blame. History shows exactly where this three-step pattern leads.

Monday: Trump tries to fire Fed Governor Lisa Cook (first time in 111 years). Thursday: CDC chief dismissed, four top scientists resign. Discredit, dismiss, blame. History shows exactly where this three-step pattern leads.

This Friday NEMI 2025 is at Northeastern in Boston, 8 talks, 24 roundtables, 90 posters; 200+ attendees. Thanks to goodfire.ai/ for sponsoring! nemiconf.github.io/summer25/ If you can't make it in person, the livestream will be here: www.youtube.com/live/4BJBis...

Announcing a deep net interpretability talk series! Every week you will find new talks on recent research in the science of neural networks. The first few are posted: jackmerullo.bsky.social, Roy Rinberg, and me. At the @ndif-team.bsky.social Youtube Channel: www.youtube.com/@NDIFTeam

The New England Mechanistic Interpretability Workshop, NEMI 2025 is August 22 in Boston. Talks, posters, meals, discussion... Most of all, an excellent chance to chat about new ideas with other great researchers in the field! Help spread the word - register and repost - bsky.app/profile/koy...

The new "Lookback" paper from @nikhil07prakash.bsky.social‬ contains a surprising insight... 70b/405b LLMs use double pointers, akin to C programmers' double (**) pointers. They show up when the LLM is "knowing what Sally knows Ann knows", i.e., Theory of Mind. bsky.app/profile/nik...

FRIENDS: American science is being decimated by Congress NOW. Your help is needed to fix this. The current DC plan PERMANENTLY slashes NSF, NIH, all science training. Money isn't redirected—it's gone. Please read+share what's happening thevisible.net/posts/004-s...

Because of propaganda Americans do not understand what Rubio is doing with visas. "I gave you a visa to come and study," they think. x.com/CitizenFree... NO, he has not!! Please help explain to X how Rubio has stopped *ALL* student visas, and how it is killing US science.

When setting up my AI lab I faced a choice between Toronto and Boston. I chose Boston, my home and the world's best incubator for research talent. Here you can take a short stroll to meet with top minds in hundreds of fields from AI to astronomy, batteries to biotech.

Black Box, Blood Money Friday evening, an Italian tourist escaped a torturer in Manhattan who was after his crypto password. I asked Anthropic's Opus 4 to analyze and explain what the episode might teach us about AI. It critiqued my guidance, instead proposing a focus on VCs:

Please join me in celebrating the contributions of our international students, researchers, and visitors. Here is a reminder of what makes America unique, and why no other nation can touch USA's 420 Nobel prizes:

How to build AI leadership in the U.S? It is not about the chips. It is about the people! I spoke about AI interpretability at ntird.gov/ last week. (NTIRD is the joint program between 23 federal agencies that coordinates government technology investments.)

My grandfather was a WW2 American Army veteran who became a proud cataloger at the Library of Congress. As a toddler I remember walking to his Library office from his A Street home, with a stop at the playground. The LOC has always been the jewel of America for me.

Leon Bottou's post ICLR thoughts are worth a read. He reminds us that modern AI is not just a product: it is a scientific wonder that we still do not understand. leon.bottou.org/news/two_les...

In academia, we treat too many of our customs as "awards to be won" rather than obligations to the community. It seems wrong that "getting a paper through peer review" is seen by many like an award, a game to win, rather than a duty. Peer review forces writers and readers to *teach* each other.

Insightful advice from M Gessen. www.nytimes.com/2025/04/21/o... Universities currently focus on competition—getting as many applicants as possible… real estate… endowments… rankings in the US News & World Report. They need to set all of that aside and focus on teaching as widely as possible.

Government research programs like #NAIRRPilot are very different from industry efforts. Because just making smart technology products is not enough for leadership: We must nurture the community of human experts. And government builds those communities from the ground up ↘️

"we must recognize that our superpower as academics is that nobody is in it for the money"

Credibility, not capability. The most important thing we build in technology and academia is not capability, but credibility. It does not matter how fast we calculate, how smart we are, or the number of products or papers we make, if we cannot answer "Why should anybody believe anything we say?"

Very cool work that introduces "concept heads" that copy meanings from one token to another. I love that our activation-based analysis of multilingual representation has now additional insight from weight space analysis: bsky.app/profile/sfeu...

Sheridan asks whether the Dual Route Model of Reading that psychologists have observed in humans also appears in LLMs. In her brilliantly simple study of induction heads, she finds that it does! Induction has a Dual Route that separates concepts from literal token processing. Worth reading ↘️

In my life I have paid a lot of tax. And every year, after participating in debates about how to spend it all—town, state, and country—I have been proud to write each tax check even when I disagree with the decisions. This is the first year I have had serious misgivings.

The need for interpretability to unlock innovation is not new. Contrast biology before and after Nirenberg 1962. Solving **interpretability**, decoding the mysterious information encoded in DNA, is the key to innovation in modern biology. bsky.app/profile/rod...

The effects of transparency and interpretability in the AI innovation ecosystem are already visible today. EliahuHorwitz arxiv.org/abs/2503.10633 HF viz shows it: the light blue explosion of SD innovation is from interpretable methods like @rohitgandikota.bsky.social Sliders sliders.baulab.info

What will be the linchpin for AI dominance? Read our NSF/OSTP recommendations written with Goodfire's Tom McGrath tommcgrath.github.io, Transluce's Sarah Schwettmann cogconfluence.com, MIT's Dylan Hadfield-Menell @dhadfieldmenell.bsky.social TLDR; Dominance comes from **interpretability** 🧵 ↘️

Here is a litmus test for you. What is your choice? As AI becomes more powerful and begins to act with social awareness, with a theory of mind, then a theory of self, and beyond that when it begins to exhibit traits of consciousness, then what?

Today we launch a new open research community It is called ARBOR: arborproject.github.io/ please join us. bsky.app/profile/ajy...

DeepSeek R1 shows how important it is to be studying the internals of reasoning models. Try our code: Here @canrager.bsky.social shows a method for auditing AI bias by probing the internal monologue. dsthoughts.baulab.info I'd be interested in your thoughts.

Can an AI make a painting with no paintings in its training data? huggingface.co/spaces/rhfei... This new demo of @rhfeiyang.bsky.social and Joanna Materzynska and @rohitgandikota.bsky.social's Art-Free Diffusion is worth a spin. You can use AI models that have seen no art, or just specific art!

⏳ Final call! Applications for NDIF's 405b pilot program close tonight! Don’t miss the opportunity to experiment with Llama-405b and shape groundbreaking AI research infrastructure. Details below 🧵⬇️

What was the most important machine learning paper in 2024? My Famous Deep Learning Papers list (that I use in teaching) does not include any new ideas from the last year. papers.baulab.info Which single new paper would you add?

I'm at #NeurIPS2024 Safe and Trustworthy Agents Workshop (West bldg Ballroom C) and will give a talk this morning at 9:40AM. Will talk about interpretability surprises and challenges, and respond to @ilyasut.bsky.social's thoughts on complex agentic AI. Come join!

Apply to the NDIF Summer Engineering Fellowship to come to Boston to work on the National Deep Inference Fabric, a cutting-edge infrastructure for research in large-scale AI. A really interesting project that crosses ML, HPC, Systems, PL, UX/Viz, research, and engineering. ndif.us/fellowship.h...

The Phase 2 NDIF Pilot is open for a short window. Apply now to get research capacity on Llama 405b. Deadline is December 31. It is not easy to crack open 405b for research, but NDIF solves the key engineering problems for you. Phase 1 powered several very interesting ICLR submissions...

PhD Applicants: remember that the Northeastern Computer Science PhD application deadline is Dec 15. It's a terrific time to do a PhD, with so many interesting things happening in AI. Apply here: www.khoury.northeastern.edu/apply/phd-ap...

Learning style from a single image is difficult, but what if you had access to an **image pair** instead? I’m excited to share our #SIGGRAPHASIA2024 work PairCustomization, on customizing text-to-image models with a single image pair!! project page: paircustomization.github.io 1/3

Do you need to copy art to make art? Hui Ren's and Joanna Materzynska's Art-Free Diffusion tests this question and lets you make "imitation-free" AI art Github: github.com/rhfeiyang/ar... Arxiv: arxiv.org/abs/2412.00176 Website: joaanna.github.io/art-free-dif... X: x.com/materzynska/...

It's been a year since the last post. Now just back from EMNLP/BlackboxNLP 2024, which was as excellent and as interesting as last year. A journalist asked me "how quickly is interpretable ML changing"? My answer: We learn big new things about AI internals every month. It is moving very quickly.

Headed to Singapore for CoNLL, BlackboxNLP, EMNLP. Looking forward to learning about your interpretability research! Reach out if you'll be there and you'd like to chat.

Stuck on what Thanksgiving dish to cook? 🦃 Here are some AI generated ideas - by composing concepts like "cooked" and "fancy" dinner. 🍗🥧 You can also blend and explore “vegetarian”, “healthier diet”, and many more! Stay tuned for technical details! 🧵 Happy Holidays!🎄✨ @davidbau.bsky.social