Title: Big hypercubes, or: how to collaborate on math with a machine
Abstract: I'll tell a story about a recent project I worked on with a bunch of combinatorists and a mathematician at Google DeepMind, in which we used a protocol called AlphaEvolve to help us work out an interesting example. I won't give any proofs from the paper, but everything in it is completely elementary and could have been observed 50 years ago, but happened not to be. I'll explain how AlphaEvolve works, what the iteration between mathematicians and machine learning tools looks like in practice, and how this paper inspired the following variant of Sudoku:
https://dyadiku.netlify.app/
Reza Gheissari
Email