How can deep learning help pure mathematicians?
Appears in collection : FLAIM: Formal Languages, AI and Mathematics
The last few years have seen the first applications of machine learning in pure mathematics. I will survey some of these, including recent work with the DeepMind team on Bruhat intervals. All applications that I know of so far are tentative. My sense is that we are beginning to explore a rich world, but I could certainly be wrong. I will try to explain the main lessons that I have learnt and where we might hope to see interesting results in the future.