Luiz Pessoa

*The Entangled Brain* is out now! I wrote it with the general reader in mind, so if you want to learn about the brain from a fresh perspective give it a try. I'll write some threads with some of the main themes soon 🧠 mitpress.mit.edu/9780262544603/…

1/N Long horizon, complex tasks that truly matter in everyday life are not solved problems by today’s robotics, requiring planning, object detection, object manipulation, and failure recovery. That's why Stanford's BEHAVIOR Challenge is back for year 2! Last year, the winning solution reached only 12.4% full task success. This year, the BEHAVIOR challenge has more tasks, better evaluation, and is easier to use. 🚨 ⏰ Submission deadline: 10/16/2026 📣 Winners announced: 11/04/2026 🏆 Prize pool: $11,000

Function is not tied to specific brain structures or time scales. The brain is not just a set of parts that each “do their own job”. Instead, networks are multifunctional and active across different speeds simultaneously. doi.org/10.1073/pnas.2… #neuroscience

@MillerLabMIT Earl, you're not one for understatements, but this time, yes!

This seems like a bold claim, given that we still don’t understand how consciousness works and current evidence suggests the brain operates differently from AI. A Silent Workspace In Claude Mirrors Key Features of Human Consciousness m.slashdot.org/story/456104 #neuroscience

Don't forget our Salon discussing 𝘄𝗵𝗮𝘁 𝗯𝗿𝗮𝗶𝗻𝘀 𝗱𝗼 𝗮𝗻𝗱 𝗱𝗼 𝗻𝗼𝘁 𝗱𝗼... July 9. Open to all but you need to register (zoom account is necessary)

こちらアップデートしました kamitanilab.github.io/stats-math-sup… 因果推論と機械学習の初歩を追加したほか、全面的に行間を埋める編集を加えました

In 1964 C. P. Willans wrote down a single closed-form expression that returns the n-th prime number exactly. The formula is perfectly exact — and gloriously useless to compute — but it settles a quiet question: yes, such a formula exists.

Before going full cyborg and replacing all your neurons with artificial units, it's good to think about the consequences. Yogi Jaeger on AI vs life and mind. braininspired.co/podcast/241/

This paper highlights one of the greatest failures in modern scientific publishing. At top journals, papers routinely spend 2–3 years in peer review. Those years consume millions of dollars in grant funding and years of student and staff labor performing experiments that were never part of the original scientific question. Yet the central scientific claims barely change. The reason is that journals abandoned editorial responsibility. Instead of evaluating whether a discovery was important, editors substituted the volume of work for scientific merit. They built a system where the labs willing to produce the most additional data were rewarded, regardless of whether those experiments materially changed the paper. The incentive was obvious: corporate publishers wanted to derisk editorial decisions. Rather than relying on scientific judgment, they demanded overwhelming amounts of data so responsibility shifted away from the editor. Peer review also changed fundamentally. Editors were once neutral arbiters between reviewers and authors. Today, editors frequently align with reviewers in demanding years of additional work while authors have little choice but to comply. AI has the potential to make much of this obsolete. If the central claims are already evident in the preprint, there is no justification for burning years of scientific talent and millions of taxpayer dollars generating figures that do not meaningfully change the conclusions.

@JulioMTNeuro Agree, not mutually exclusive. Overall I'm more interested in multiregion circuits so I'll bet on key computational properties emerging from distributed circuits. With notable contributions from local properties ofc. (But, yes, mostly a matter of where one puts most weight.)

@PessoaBrain They are not exclusive views. Neurons in PFC are larger but fewer than V1, they have more NMDA and spines. Basically the ability to connect depends on morphology and intrinsic properties. That includes interneurons. Circuit computations and cell types are linked.

Is human intelligence based on new neuron types or how they act together? Empirical question of course but I'm betting on more of the latter.

🚨 DEADLINE EXTENDED 🚨 Due to high demand & #CCS2026 synergy, submissions stay open until July 22 (AoE)! Got a breakthrough in decentralized computation or collective behavior? Submit your abstract & get published in our @SpringerNature proceedings. 👇 guided-self.org/gso-2026.html

@JulioMTNeuro Interesting, my view is a bit different. Mostly about, reciprocal, reentrant, nonlinear circuits. Sure, new neuron types might help too.

@PessoaBrain It is more about computational capacity of single neurons and how they contribute to network processing. The complexity of the transfer functions matter. Cognition is not about adding more of the same hardware and connections but has that other aspect. IMO.

Learning coordination

**Hacking Collective Behavior*" meeting at Politecnico di Torino, Italy Engineering Collective Behavior by Mario Di Bernardo

Found this if you read Italian!

@PTenigma 19: Ogni abuso sarà punito

2 weeks ago at OHBM it was a good opportunity to speak French, parce que c'est l'occasion. I wanted to thank Hervé for finding us a great venue for our ENIGMA meeting and the taxi driver helped me to pronouce it correctly. French is a very interesting language as people think of it as an expansive language, but look at its very high compression: look at the example "Tout abus sera puni". I think this cannot be said in any other language in 16 letters?

More evidence for a causal role of brain rhythms in cognition. Analyzing different brainwave frequencies shows how the brain mixes multiple sources of information when forming decisions. doi.org/10.1371/journa… #neuroscience

Excellent piece about AI and the future of work. (congratulations on such a clear and cogent piece)

A short Research Briefing on our recent work: “How the zebrafish brain weaves recent experiences into future decisions” Grateful to the Nature team for helping tell the story clearly. nature.com/articles/d4158…

This looks absolutely superb