Happy New Year! Here's some microorganism porn. vimeo.com/1024357218 (srsly though. it's mesmerizing.)

claude code is fucking insane i know literally NOTHING about Hegel. ZERO. and it just built me a complete system of German idealism

Fascinating story: 1. Fear-mongering about peanut allergies multiplies in the late 90s. 2. AAUP advises "no peanuts til age 3", despite NO evidence. 3. This advice CAUSES a massive spike in peanut allergies! 4. They had to do a large study just to undo the damage & convince everyone it was safe!

Astonishing: the default behavior of Chrome's "Listen to this page" feature is NOT to simply read you the page, LIKE YOU WOULD EXPECT, but instead to make one of those goofy NotebookLM AI podcasts... and I had to search the Help docs to figure out how to change it. Thanks, Google.

MAORL (Multi-Agent Olof RL) arxiv.org/abs/2512.16705

This is a travesty!! Also I have questions... for the branches that aren't terminated, where will they move? Will they be in the same building, under a different name? In which case, will it all be functionally the same thing (but with lots of political disruption)?

Random observation: did you know this?! I knew about how Einstein's general relativity is required to make GPS navigation work, but I'm not as familiar with how quantum mechanics factors into semiconductor technology. Remarkable.

In applying AI to material science, biology, etc, capitalism is trying to shed science. The point is to substitute the engineering of a machine that can generate what science has hitherto done, but without having people know things. Knowledge ultimately residing in private property is the dream.

[In considering candidates] "I actually view publication volume completely negatively... I've learned, from experience, that those with a really high volume [don't have a depth of knowledge]." —@abeirami.bsky.social #slowscience FTW

I thought this was a fascinating way to motivate model merging: to be able to *interpolate* between model classes, rather than be forced into discrete choices. (models trained on different datasets, models of different sizes, etc.) "Interpolation and alignment are two sides of the same coin."

Did you know that you could create hybrids across different neural net architectures—*without* updating the weights?? Come check out my poster at the UniReps @unireps.bsky.social workshop at 3:45pm today at #NeurIPS! (Or you can already browse the posters on the wall in 20D throughout the day.) 😃

Join us this Saturday at @neuripsconf.bsky.social 2025 for the @unireps.bsky.social Workshop.

Me, at people with #NeurIPS badges walking away from the convention center right now: "Don't they know that Rich Sutton is speaking in 10 minutes??? 😨"

🔵🔴 Join us for the UniReps Workshop: Unifying Representations in Neural Models at @neuripsconf.bsky.social 2025! 📍 Ballroom 20D, San Diego Convention Center Dec 6 Don’t forget to fill out the participation form. Joining in person or remotely? We welcome your questions for the panel. 🔗 unireps.org

So stoked to be going to my first #NeurIPS!! It’s crazy that in 10+ years of robotics and AI, I've never been to the great Lollapalooza of Machine Learning. 🥳🎆 I’m presenting my Frankensteinian efforts to stitch together parts of different neural networks! 🧪

Check out our new work on using low-rank perturbations to make evolution strategies work for billion-parameter models.

The UniReps Workshop accepted papers are out! 🎉 Huge thanks to the authors, reviewers, and incredible AC committee for their dedication and effort in making this happen. 🙏 openreview.net/group?id=Neu...

What if instead of buying software from a store, you could grow it in your garden?

What an incredible tool! I think I'll be returning to this a lot over the next month! The most interesting papers might be the ones that seem misclassified or out of place...

It's too easy to write garbage survey papers and get lots of citations. Over time, I've learned to ignore most survey papers and recognize some markers of quality.

The blog post is available: blog.arxiv.org/2025/10/31/a...

It's been 8 months since I notified @springernature.com of a clear-cut case of fake references in a book chapter which they published. They have still not taken any action as they claim to still be investigating. I will not review for Springer again until this matter is satisfactorily resolved.

UBC’s department of Computer Science invites applications for up to 2 full-time tenure-track positions. The department is particularly interested in researchers in: visualization, robotics, reinforcement learning, data management, and data mining. @cs.ubc.ca science.ubc.ca/about/careers

Translation: You have no new messages.

This year's ALICE guest speakers 🧑‍🔬 - Angel Goñi-Moreno - @angelgm.bsky.social - Alyssa Adams - @alyssa-m-adams.bsky.social - Alexander Mordvintsev - Eric Medvet - @ericmedvetts.bsky.social - Kyrre Glette - @kyrre2000.bsky.social - Stefano Nichele - @stenichele.bsky.social - Susan Stepney

"Consider another human difference: financial solvency, which can be measured and quantified, just like IQ. It is heritable in twin studies and (less so) in SNPs. But would you bet that sooner or later we are going to know 'what’s going on' with your bank account at the level of genes?" Same for IQ.

GPT now has ads. 💀 They are voluntary... for now.

Q. Who aligns the aligners? A. alignmentalignment.ai Today I’m humbled to announce an epoch-defining event: the launch of the 𝗖𝗲𝗻𝘁𝗲𝗿 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗼𝗳 𝗔𝗜 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗖𝗲𝗻𝘁𝗲𝗿𝘀.

At #AutoML25, Dr. Manuela Veloso calls for 3 major directions for AI research: 1) AI must know what it doesn't know (uncertainty). 2) AI must be able to continually improve (continual learning). 3) AI systems must include humans at all times to be more robust (automation is always only partial).

No mires pabajo

Merely a day after I call out OpenAI for being so closed (bsky.app/profile/ntra...), they release an open weights GPT. Glad they're finally reading my skeets! 😏

With its focus on short term profitability, the US has forgotten how to make markets competitive. And companies have forgotten that they are only *vehicles* for the final product, not ends in themselves.

I’m very excited to announce that I’ve just signed a contract with @princetonupress.bsky.social for a new book, tentatively titled “The Genomic Code” 📖 😊

In reality, all environments are only partially observable, so this is the only regime in which RL should be evaluated.

Send like a useful framework. Looking forward to reading this new position paper.

I just realized that all these hyped-up “AI scientist” concepts all concentrate on the idea of *discovery*—new algorithms, new materials, new products—without exception. Not a one of them focus on *understanding*—arguably the larger role of a scientist! 🧐

Proud of our team's talks and posters at #IC2S2 in Norrköping Sweden - what a great week. We're excited to host next year in Vermont!

We are going to spend the next few years finding out exactly *why* it was a horrible idea to unleash a tsunami of vibe-coded apps made by idiots and scammers on an unprepared world. Welcome to the Entirely Foreseeable AI Consequences Era.

Apparently Papers With Code was abruptly sunsetted for reasons that are unclear. 😢

😞 So sad to see paperswithcode is discontinued – but grateful that the @hf.co team is, as always, stepping up to support the community! It was an incredible resource for use cases, common themes in papers, and visualizing how models have improved on evals over time: huggingface.co/papers/trend...

Been studying Mixture of Experts recently. It's baffling to learn that typically different tokens are routed to *different* experts. How can you be an "expert" at a particular word?? I suspect this name lends us very poor intuition for what these models actually do! (shows 2 tokens and 4 "experts")

There are people, in tech (and now in the government!), who will mislead you about what current AI models are capable of. If we don't call them out, they'll drag us all down.

more evidence for “a bunch of AI cultists dismantled large parts of the federal government because they believed that their LLMs were intelligent and could do the work better” thesis

Get $10-25k to develop a piece of software that benefits the scientific community. Requirements are pretty broad and they have not gotten as many applications as they'd like. But you must apply in the next 2 days!

Nice perspective on Research Scientist jobs in AI industry in recent years. newsletter.ruder.io/p/thoughts-o... Helpful for considering academia vs. industry. The opportunities for academic freedom in industry are far more limited than when I first started considering academia ten years ago.

Really loved this blog post by my labmate, Nate. Did you know that basically all multicellular organisms reproduce sexually?? I didn't. This leads to some compelling consequences. Maybe sexual reproduction is what makes multicellularity possible?? thinkingwithnate.wordpress.com/2025/07/01/b...

I designed an experiment using the concepts of intelligence as described by @paulstamets.bsky.social and @drmichaellevin.bsky.social. Is tap water intelligent? 6 minutes is all you need to find out: youtu.be/YVt-QHudRDA?...

1/3) This may be a very important paper, it suggests that there are no prediction error encoding neurons in sensory areas of cortex: www.biorxiv.org/content/10.1... I personally am a big fan of the idea that cortical regions (allo and neo) are doing sequence prediction. But... 🧠📈 🧪

Not really an overstatement to say that the test of a free society is whether or not comedians can make fun of the country's leader on TV without repurcussions.

This is part of a larger question I'm stumped by. Why have pop-ups become "okay"? Even reputable businesses constantly throw unbidden things in front of the UI I'm actively trying to use. Is there really no other way to bring attention to new features / other services provided by the business??