Stanislas Dehaene @standehaene.bsky.social

Chercheur en neurosciences cognitives

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Mon nouveau livre "Le rectangle de Lascaux : et Homo sapiens inventa la géométrie" sort le 28 janvier. J'y aborde l'origine des mathématiques depuis la préhistoire jusqu'à nos jours ; c'est un peu la suite de mon tout premier livre, "La Bosse des Maths". Explications:

[Nick Blauch @ ICML 2026 🇰🇷 @nmblauch]

Machines see the world fundamentally differently from humans, and it's costing us a lot of tokens.

Where machines see uniform rectangles, our brain sees a variable resolution retinal array reformatted into a magnified manifold in primary visual cortex, enabling high-resolution and wide field-of-view at minimal computational cost.

In our new ICML paper, we enable the same mechanism for deep vision models.

🧵 ↓

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

@kanair I agree. The virtue of the J space, and the global neuronal workspace it started from, is that they are precise enough to be testable and lead to scientific progress. Once these mechanisms are understood, ill-defined issues of “phenomenal” awareness and “qualia” will dissolve

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Excellent summary of gorgeous work by Jean-Rémi King!
Still, the J-space is something quite different and a genuine discovery in my opinion: a selective capacity-limited system with quite restricted content, unlikely the massively parallel machinery for vision or word processing

[Jean-Rémi King @JeanRemiKing]

Excellent investigation of the mechanics of language models.

🫢A notable blindspot, however: several teams have actually been **directly** comparing the working of LLMs to those of the human brain 🧠 for a while.

Here is a thread to highlight some of our (and our competitors'!) findings 👇

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

One of Gurnee et al's experiments, which is a dream for neuroscientists, consists in swapping J-space contents and watch the model's reasoning change accordingly. Strikingly, only high-level non-routine behavior is affected, while routine tasks remain unchanged (Figure 20).

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Lionel Naccache and I wrote a commentary on this work from a neuroscience perspective. You can read it here :
https://t.co/oaoqtPPSif

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

@davidmanheim Absolutely :-) Very exciting

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

The global neuronal workspace (GNW) is currently the best documented neuroscience mechanism by which conscious processing arises in the human brain — and now Anthropic researchers have discovered a similar workspace inside their large language model !

[Anthropic @AnthropicAI]

New Anthropic research: A global workspace in language models.

Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.

We found a strikingly similar divide inside Claude.

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

@VictorShestakov No, actually we wrote it over the past few weeks, and had several rounds of dialog with @Jack_W_Lindsey which led to several new tests of the similarity between the LLM's J space and the human Global Neuronal Workspace.

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Lionel Naccache and I wrote a commentary on this work from a neuroscience perspective. You can read it here :
https://t.co/oaoqtPPSif

[Anthropic @AnthropicAI]

New Anthropic research: A global workspace in language models.

Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.

We found a strikingly similar divide inside Claude.

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Many thanks to @Jack_W_Lindsey and the @AnthropicAI team for their great research and for several weeks of intense discussions that led to new experiments

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Lionel Naccache and I wrote a commentary on this work from a neuroscience perspective. You can read it here :
https://t.co/oaoqtPPSif

[claudine monteil @claudinemonteil]

Mon père, le #mathématicien Jean-Pierre Serre, #médailleFields à 27ans, professeur au Collège de France à 29ans, prix Abel notamment, ne supporte pas, comme ce prix Nobel français de physique, que les journalistes se vantent d'avoir été nuls en maths et en physique. @cdf1530 @abel_prize

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

In this new JEP:General paper, we show that geometric shapes are organized as tree structures in a language of thought.
We apply several linguistic tests for nested constituents, including structural ambiguity, constituent subparts, and syntactic movement.
https://t.co/LZrECP66ww

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Exciting work! AI is exciting, but real neurons can do much more than artificial ones.

[Ido Aizenbud @IdoAizenbud]

TL;DR: A single cortical pyramidal neuron is not a point; it's a powerful, noise-robust, general-purpose computational unit.

Dendrites are not just wires to help neurons connect; they are the substrate for complex nonlinear computation. (15/15)

preprint: https://t.co/m2a8ieuo8A

[Jared Duker Lichtman @jdlichtman]

Last year, this news would have been science-fiction:

GPT5.4 Pro found an elegant solution to a 60 year old conjecture, Erdős Problem #1196. The proof subverted the natural human intuitions.

A day later, the proof was fully formalized in Lean by Gauss.

What a time to be alive

[Jean-Rémi King @JeanRemiKing]

🧠 the Digital Brain Project is now live:

$5M total · up to $500k per selected team

Let's open-source the modeling of the human brain brain activity!

➡️Apply on: https://t.co/W4HFA5PQBX

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Une petite vidéo très accessible sur quelques-unes des grandes idées de mon livre récent : « le rectangle de Lascaux »

https://t.co/y6goYf0pe7

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

A new paper from the lab!
We use MEG and a "local/global" design in the language domain to ask whether the transitions between words in a sentence are encoded by a shallow transition-probability mechanism, in parallel to a tree-based syntactic mechanism.
https://t.co/igdWANzUbY

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

@doristsao @dileeplearning @WadiaVarun @UeliRutishauser Congratulations for a fantastic study! Are all object axes equally present in imagery, or can you see if higher-level, more abstract ones are somehow privileged ?

[VarunWadia @WadiaVarun]

1/8 Our preprint is now a peer-reviewed paper :) Big thanks to our reviewers who pushed us to examine our results more carefully and Olivier Wyart (https://t.co/pQgGhUgdQi) for the exquisite visual. https://t.co/uQzvMXhB7r

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

https://t.co/9ETaP8dv4U

[Stanislas Dehaene @standehaene.bsky.social @StanDehaene]

Je donnerai prochainement une intervention à la BNF sur l'imagerie cérébrale dans le cadre de cette conférence :
"De la trépanation à l'imagerie cérébrale : révolutions dans l'exploration du cerveau" https://t.co/Dtow38EQ7d