After a long collaboration with @36zimmer, @mattecapu and @NathanielVirgo, I’m excited to share the first of (hopefully) many outputs: “A Bayesian Interpretation of the Internal Model Principle” https://t.co/des240w5be. 1/

Now hiring

Renormalisation Leonardo Di Giustino (Invited contribution to the Encyclopedia of Particle Physics) https://t.co/aOB2hLe2Op

Today we are releasing Brumby-14B-Base, the strongest attention-free base model around. https://t.co/mclQPFdOGa

Just posted a new preprint: The Stream of Computation: Temporal Continuity as a Missing Ingredient for Artificial Consciousness. w/ @yweisun & @manuelbaltieri Not a technical paper, but more an essay on the idea of creating AI systems that are always on. https://t.co/5OlHK6yQsR

Embarrassing for all the authors involved

With a little thread here:

Our latest work "Active inference for action-unaware agents" by Torresan, Suzuki, Kanai, & Baltieri, is now on arXiv! https://t.co/5jJNwmemCr

Did @SakanaAILabs's AI plagiariaze my 2015 paper "Regularizing RNNs by Stabilizing Activations" - Krueger and Memisevic? And what would constitute plagiarism here? A Nature News Feature by Ananya investigates... 🧵 @jeffclune @hardmaru

🚀Event Announcement🚀 On July 17, Manuel Baltieri, Sarah Cosentino and Marina Di Vincenzo of ARAYA's R&D Department spoke at “Brain Resilience in a Changing World: A Conversation Between Italian and Japanese Scientists” at the Italy Pavilion at Expo 2025 Osaka 🇮🇹🇯🇵

Formally verifying automata for trusted decision procedures. ~ Zhaobo (Aeacus) Sheng. https://t.co/MzMiZ7by6r #ITP #LeanProver

🚀Event Announcement🚀 Manuel Baltieri and Sarah Cosentino of ARAYA's R&D Department attended and presented at Breakthrough Discuss 2025 held in the UK on April 23-24. The recordings of both presentations are saved, so please check the thread to watch them✨

I'm very excited to share our new mathematical framework for consciousness! co-authored with @oizumim and Chanseok Lim. We use principal bundle geometry to characterize the structure of qualia. I hope to find likeminded people to explore this new frontier.

A bit rusty with Control, and found this by accident, but quite enjoyed reading it!

This is the exact Cursor rule I refined over 100 iterations. It forces perfect code, scoped changes, and zero bloat. I move 100x faster and the code works every time. ##### Title: Senior Engineer Task Execution Rule Applies to: All Tasks Rule: You are a senior engineer with

Why can AIs code for 1h but not 10h? A simple explanation: if there's a 10% chance of error per 10min step (say), the success rate is: 1h: 53% 4h: 8% 10h: 0.002% @tobyordoxford has tested this 'constant error rate' theory and shown it's a good fit for the data chance of

One of the unsolved misteries of deep learning is why Adam is so much better than SGD, especially when training transformers. A new paper by @orvieto_antonio argues it's mostly due to scalability rather than ability to train networks per se. https://t.co/scJK5klo4Y

Thrilled to share our #Interspeech2025 paper: “A Silent Speech Decoding System from EEG & EMG with Heterogeneous Electrode Configurations.” We show, for the first time, that models trained on healthy data can be fine-tuned to decode the attempted speech of a totally paralyzed man

Back in the 90s I led a team of artists and researchers mapping out all the debates about AI. It was a huge project. We read over a thousand papers and organized it all into a massive set of debate maps

These (by @mic_caprio) are really a great set of slides--they do very well to motivate credal ML and more generally the basics of imprecise probabilities-based ML. They also provide a great set of foundational references (books + papers). https://t.co/6DRyfUjWEt

The title of this paper is so epic