Mathematician (algebraic geometry, motives & friends, singularities in statistics and ML).
'Geometry is successful magic' (R. Thom)
University of Essen.
XKCD today https://t.co/69oWYIlgko is totally right. Being aware of the geometry around us and that the sun moves 15 degrees per hour has helped me select seats many times.
There's this troublesome pattern where OpenAI employees seem to think that it's okay to handwave what the law says, and replace it with something else that they think the law _should_ say.
Let me publicly express my exasperation with people who write math papers giving a non-obvious definition of something, and don't take the time to explain what would have happened if you had taken the “obvious” or “naïve” definition. 🧵⤵️ •1/19
Introducing Lattice Deduction Transformers: An 800k-parameter looped transformer that reasons like a SAT solver achieves 100% on Sudoku-Extreme with only 15 minutes of training.
A collaboration between @axiommathai, @AmherstCollege and @BarnardCollege.
@ArthurSteggles@prz_chojecki Because the definitions themselves rely on rather difficult theorems, which rely themselves on sophisticated definitions, which...
In cohesive homotopy theory,
the formula for the ℝ-rational homotopy type
of mapping spaces out of a smooth manifold
has a one-line proof:
https://t.co/dOoJuzZGLq
Here's a "theorem" in the complex variables section proved by sorry: "If g is complex differentiable at every point with Re z \geq 0, then g is complex differentiable everywhere." This is then used in the proof of another theorem in Lecture21.lean. All of this is marked as a "pass".
Can LLMs discover new laws of physics?
We present DiscoverPhysics, a pipeline to benchmark LLM agents on experimentation, analysis and discovery.
https://t.co/FSjixhuwc6
Co-led by @LindsayMSmith3 w/ Peter Melchior, @kdqg1@andrewgwils@Pavel_Izmailov Carol Cuesta-Lazaro 1/10