Version 2 of Theory of Contravariance w/ @dyamins.bsky.social is out! New material on contravariance for Transformers, and the theory of Representational Similarity Analysis (RSA) and centered kernel analysis (CKA)/Procrustes.

Aran Nayebi@anayebi.bsky.social · 3w ago

1/6 Why have deep neural networks aligned so strongly with brains for the past 15 years? What explains it? @dyamins.bsky.social & I make progress on this question in our new paper👇 In a nutshell, we *prove* that for sufficiently hard tasks, the choice of alignment metric does *not* matter.

Now accepted to the AI, Ethics, and Society 2026 conf under the new title: "When Do AI Gains Become Broadly Shareable?" With all the rapid progress in AI, it is critical to now rigorously characterize the institutional conditions that allow gains from greater AI capability to be broadly shared.

Aran Nayebi@anayebi.bsky.social · last yr.

Can a Universal Basic Income (UBI) become feasible—even if AI fully automates existing jobs and creates no new ones? We derive a closed-form UBI threshold tied to AI capabilities that suggests it's potentially achievable by mid-century even under moderate AI growth assumptions:

If anyone ever asks you: "What has AI/NeuroAI taught us about the brain?" Here's a (woefully incomplete!) list of what you can say:

Why, according to Terry Tao, the recent disproof of the Jacobian conjecture in N > 2 dimensions was not mere brute force and required what many mathematicians would characterize as "creative insight" had a human come up with it!

The blogpost exposition on our "zippering theorems" in Section 4 from our recent Contravariance Theory paper is now out! These zippering theorems explain why end-to-end optimization for downstream tasks often seems to strongly constrain upstream representations.

Aran Nayebi@anayebi.bsky.social · 3w ago

1/6 Why have deep neural networks aligned so strongly with brains for the past 15 years? What explains it? @dyamins.bsky.social & I make progress on this question in our new paper👇 In a nutshell, we *prove* that for sufficiently hard tasks, the choice of alignment metric does *not* matter.

And come see the substack series: danyamins.substack.com/p/the-theory...

The Theory of Contravariance, Part 0

Breaking down our extensive new theory paper.

danyamins.substack.com

Aran Nayebi@anayebi.bsky.social · 3w ago

1/6 Why have deep neural networks aligned so strongly with brains for the past 15 years? What explains it? @dyamins.bsky.social & I make progress on this question in our new paper👇 In a nutshell, we *prove* that for sufficiently hard tasks, the choice of alignment metric does *not* matter.

1/6 Why have deep neural networks aligned so strongly with brains for the past 15 years? What explains it? @dyamins.bsky.social & I make progress on this question in our new paper👇 In a nutshell, we *prove* that for sufficiently hard tasks, the choice of alignment metric does *not* matter.

Had a blast presenting our virtual zebrafish paper at #FENS2026 in Barcelona in the "Computational Astroscience" symposium -- the first autonomous agent that can predict *whole-brain* neural-glial data!

Reece Keller@reecedkeller.bsky.social · last yr.

1/ I'm excited to share recent results from my first collaboration with the amazing @anayebi.bsky.social and @leokoz8.bsky.social ! We show how autonomous behavior and whole-brain dynamics emerge in embodied agents with intrinsic motivation driven by world models.

We're at @fens.org! #FENS2026 #FENSGlia Posters: PS02-07PM-558, PS03-08AM-495, PS05-09AM-674. Delighted to be presenting at the S44 Computational astroscience symposium on Thursday with T. Fellin, @anayebi.bsky.social & T. Papouin. Thanks @inbalgoshen.bsky.social for organizing, looking forward!

In case you want to learn more about AI safety this 4th, check out the recent recording of some of my group's work on the AI Safety Research directory! www.youtube.com/watch?v=XWQJ...

Corrigibility and the ROGUE Benchmark with Prof. Nayebi, NeuroAgents Lab

YouTube video by AI Safety Research Directory

youtube.com

UAI 2026 Camera ready up on arXiv as v3! I've written a long-form LW blogpost on how the technical aspects of this work connect to NeuroAI and to AI sentience/welfare, entitled: "What Capable Agents Must Know: Why AI Consciousness May Be an Inevitable Byproduct of Capability"

What Capable Agents Must Know: Why AI Consciousness May Be an Inevitable Byproduct of Capability — LessWrong

[No LLMs were used (or harmed!) in the writing of this blogpost!] Technical results can all be found here: https://arxiv.org/abs/2603.02491 …

lesswrong.com

Aran Nayebi@anayebi.bsky.social · 5mo ago

1/ As AI agents become increasingly capable, what must *inevitably* emerge inside them? We prove selection theorems: strong task performance forces world models, belief-like memory and—under task mixtures—persistent variables resembling core primitives associated with emotion.

Check out my friend Scott Aaronson's latest blogpost for his recollections of the Aumann conference! scottaaronson.blog?p=9875

50 Years of Aumann’s Agreement Theorem

One of the most popular posts in this blog’s history was Common Knowledge and Aumann’s Agreement Theorem, based on a lecture that I gave to high-school students 11 years ago. One of the…

scottaaronson.blog

Aran Nayebi@anayebi.bsky.social · 2mo ago

Honored to meet Nobel laureate Bob Aumann & speak on AI safety in Paris at his 96th birthday celebration: Half a Century of “Agreeing to Disagree”. Scott and I met him here for the first time. When Scott mentioned Aumann Hall at Lighthaven, Bob said: “I guess I’ve made it then!”

How did Aumann's agreement theorem come to be? I've uploaded the full video👇from Nobel laureate Bob Aumann's 96th birthday conference in Paris, celebrating 50 years of “Agreeing to Disagree.” At 1:11:09, I ask Aumann about meeting von Neumann & Morgenstern, their influence on his career,

Aran Nayebi@anayebi.bsky.social · 2mo ago

Honored to meet Nobel laureate Bob Aumann & speak on AI safety in Paris at his 96th birthday celebration: Half a Century of “Agreeing to Disagree”. Scott and I met him here for the first time. When Scott mentioned Aumann Hall at Lighthaven, Bob said: “I guess I’ve made it then!”

My AAAI 26 talk on the first formal guarantees on corrigibility and the limits of safety filters is now online: underline.io/lecture/1442...

Core Safety Values for Provably Corrigible Agents

On-demand video platform giving you access to lectures from conferences worldwide.

underline.io

Aran Nayebi@anayebi.bsky.social · 6mo ago

I'll be presenting my work *today* on the first formal guarantees addressing the decade-long open problem of Corrigibility (namely how we provably avoid loss of control with AI) in the AAAI Machine Ethics workshop (W37) at 15:15 pm ST in Tourmaline 207-209!

Honored to meet Nobel laureate Bob Aumann & speak on AI safety in Paris at his 96th birthday celebration: Half a Century of “Agreeing to Disagree”. Scott and I met him here for the first time. When Scott mentioned Aumann Hall at Lighthaven, Bob said: “I guess I’ve made it then!”

Now accepted to #UAI2026 (@auai.org)! Check out v2 of the paper on arXiv with explicit world model recovery algorithms under partial observability in Theorems 3-4. See you in Amsterdam in August!

Aran Nayebi@anayebi.bsky.social · 5mo ago

1/ As AI agents become increasingly capable, what must *inevitably* emerge inside them? We prove selection theorems: strong task performance forces world models, belief-like memory and—under task mixtures—persistent variables resembling core primitives associated with emotion.

In this era of AI doomerism, misinformation, and fear-mongering, this quote by Marie Curie is more prescient than ever: "Nothing in life is to be feared, it is only to be understood. Now is the time to understand more, so that we may fear less." Here's to understanding more.

Wow! This is a *must-watch* for anybody interested in modern AI today and where it's going, with talks from rockstars—that include 3 Turing Award winners (Dana Scott being the 4th in attendance!)—on topics such as: AI safety (Ron Rivest), using AI in research and improving reliability (Les Valiant👇

CMU MLD@mldcmu.bsky.social · 2mo ago

At Simons Institute for the Theory of Computing: The Role of TCS in Modern Machine Learning, a workshop on the interplay of theory and AI. Co-organized by MLD professor Nina Balcan and featuring Professor Tom Mitchell among many other field-defining scientists. simons.berkeley.edu/workshops/ro...

A great writeup of our virtual zebrafish agent work by @hadivafaii.bsky.social! sensorimotorai.github.io/2026/04/09/z...

Reece Keller@reecedkeller.bsky.social · last yr.

1/ I'm excited to share recent results from my first collaboration with the amazing @anayebi.bsky.social and @leokoz8.bsky.social ! We show how autonomous behavior and whole-brain dynamics emerge in embodied agents with intrinsic motivation driven by world models.

I'll be giving a talk on our virtual zebrafish at this conference -- apply now to check out the other awesome talks & speakers as well!

Janelia Conferences@janeliaconf.bsky.social · 6mo ago

📢 Applications are open! This meeting brings together scientists using data-driven #AI to analyze + model neural circuits and uncover how the brain performs complex computations. 🧠 Apply by May 28 ➡️ janelia.news/NAI26 @computingnature.bsky.social @sinzlab.bsky.social @srinituraga.bsky.social

I’ll be presenting this work (AAAI '26 Oral) in honor of Nobel-Prize-winning economist Robert Aumann's 96th birthday in Paris! Come find out why aligning to all human values is infeasible & reward hacking is inevitable! Program (w/ Steven Pinker & others!): game-theory.u-paris2.fr/WS2026-progr...

Aran Nayebi@anayebi.bsky.social · 9mo ago

We have 2 papers accepted to #AAAI2026 this year! The first paper 👇 on intrinsic barriers to alignment (establishing no free lunch theorems of encoding "all human values" & the inevitability of reward hacking) will appear as an *oral* presentation at the Special Track on AI Alignment.

Talk recordings from our CoSyNe 2026 workshop on Agent-based Models in Neuroscience are now online! Playlist: www.youtube.com/playlist?lis... RL agents and theory, interoception, biomechanical models connectome-constrained models, and more. #CoSyNe #NeuroAI #CompNeuro #RL #EmbodiedAI

CoSyNe 2026 Agents in Neuroscience Workshop - YouTube

https://neuro-agent-models.github.io/

youtube.com

Satpreet (Sat) Singh@satpreetsingh.bsky.social · 5mo ago

📣 Excited to announce the 2nd edition of our workshop “Agent-Based Models in Neuroscience: Theory, Autonomy, Embodiment & Environment” at @cosynemeeting.bsky.social #CoSyNe2026!! 🧠🤖🌍🪰🐟🐭💪🧘🏃 🗓️ March 17, 2026 📍 Cascais, Portugal 🔗 Speaker lineup and schedule: neuro-agent-models.github.io

Honored to be quoted in this great reporting by @theroberthart.bsky.social on getting the facts straight here. No, this is *not* a fly uploaded to a computer, and in fact imitation learning via RL from fly behavior was responsible for the intelligent behavior, not the "uploaded" pseudo-connectome :)

Marina Galperina@mfortki.bsky.social · 5mo ago

Viral claim on X: This is a "real uploaded animal" with “91% behavior accuracy” @theroberthart.bsky.social with @shahabbakht.bsky.social @asbates.bsky.social @anayebi.bsky.social @birchlse.bsky.social @tom-mcclelland.bsky.social: Is it though?

If you're at #Cosyne2026, stop by @reecedkeller.bsky.social's poster tonight (Poster 1-034) and ask him questions! :)

Reece Keller@reecedkeller.bsky.social · last yr.

1/ I'm excited to share recent results from my first collaboration with the amazing @anayebi.bsky.social and @leokoz8.bsky.social ! We show how autonomous behavior and whole-brain dynamics emerge in embodied agents with intrinsic motivation driven by world models.

If you're attending @cosynemeeting.bsky.social, come check out our NeuroAgents workshop on Tuesday March 17! Speakers: Omri Barak, Cristina Savin, @lilweb.bsky.social @reecedkeller.bsky.social Caroline Haimerl, Hannah Choi @xaqlab.bsky.social Srini Turaga, Yanan Sui, @trackingskills.bsky.social 👇

Satpreet (Sat) Singh@satpreetsingh.bsky.social · 5mo ago

📣 Excited to announce the 2nd edition of our workshop “Agent-Based Models in Neuroscience: Theory, Autonomy, Embodiment & Environment” at @cosynemeeting.bsky.social #CoSyNe2026!! 🧠🤖🌍🪰🐟🐭💪🧘🏃 🗓️ March 17, 2026 📍 Cascais, Portugal 🔗 Speaker lineup and schedule: neuro-agent-models.github.io

1/ As AI agents become increasingly capable, what must *inevitably* emerge inside them? We prove selection theorems: strong task performance forces world models, belief-like memory and—under task mixtures—persistent variables resembling core primitives associated with emotion.

Want to learn how to build your own biologically-plausible temporal neural networks (TNNs)? Check out the PyTorchTNN tutorial, prepared by my students @trinityjchung.com and Yuchen Shen! 👇 colab.research.google.com/drive/11QuXu... Check out the thread below for a high-level overview 👇

Google Colab

colab.research.google.com

Aran Nayebi@anayebi.bsky.social · last yr.

🚀 New Open-Source Release! PyTorchTNN 🚀 A PyTorch library for biologically-inspired temporal neural nets: unrolling computation through time. Integrates with our recent Encoder-Attender-Decoder, which flexibly combines models (Transformer, SSM, RNN) since no single one fits all sequence tasks. 🧵👇

PyTorchTNN tutorial (prepared by my students @trinityjchung.com and Yuchen Shen): colab.research.google.com/drive/11QuXu... Slides from today's talk: anayebi.github.io/files/slides...

Google Colab

colab.research.google.com

Aran Nayebi@anayebi.bsky.social · 5mo ago

Looking forward to presenting on "How behavior shapes recurrent circuits across sensory systems and species: from vision to touch" at the University of Chicago Neuroscience and ML workshop on Wednesday! Details below 👇🧵

Looking forward to presenting on "How behavior shapes recurrent circuits across sensory systems and species: from vision to touch" at the University of Chicago Neuroscience and ML workshop on Wednesday! Details below 👇🧵