TypedBio: a type-theoretic take on biology as a playground for AI reasoning. In this short article, I explored how formal structure could drive discovery. #AIReasoning #AIinScience #DrugDiscovery #ComputationalBiology #Bioinformatics.

RT @Riazi_Cafe_en: Harvard's "Dynamical Systems" (Math 118) by Shlomo Sternberg . PDF: (audio generated by Notebo….

Workshops need to be orthogonalised to eliminate redundant content.

It’s sad to see good researchers playing the influencer game. Science isn’t about “BREAKING” or “just dropped” headlines. It’s a slow, steady process that takes time. Real progress doesn’t happen overnight.

What if life’s true code isn’t written in letters, but in motion?.A short essay on dynamics, biology, and Rubel’s Universal ODE.

In the age of LLMs, figuring out what’s actually worth doing as a scientist isn’t as simple as it used to be. I think we should constantly be solving this kind of optimization problem $ argmax_f || f - AI(f) || $ where AI(f) is the part of the task f that AI can handle. The funny.

RT @pmddomingos: It’s mind-boggling that a bunch of chemical reactions can produce a human being.

Many biology experts have criticised Demis’s prediction for being unrealistic. But I like the mindset and I think it’s necessary. When you lead a group of smart extremely capable people, giving them an extremely ambitious goal is important to keep them engaged every day and for a.

RT @arashmeh: A fiction about AI, recursion, and a curve unfolding in finite time, backed by math.

I think what actually happens here is conceptually similar to diffusion models. You don't need to remember the exact position. You remember a good approximation and then bring it back to the manifold of high-likelihood chess positions. He has an internal subconscious.

RT @PhysInHistory: What is your favourite Physics book? ✍️

A fiction about AI, recursion, and a curve unfolding in finite time, backed by math.

RT @DBertsekas: I am pleased to share the full set of videolectures, slides, textbook, and other supporting material of the 7th offering of….

RT @FrnkNlsn:

How do weaker LLMs understand the rules of the tools they use? I built a small project where two LLMs play chess against each other. It turns out that even finding a valid move becomes a challenge when the model is below a certain threshold, as they don’t fully grasp the

Blindfolded with random position. the guy seems good:).