Kevin Buzzard: Can AI do mathematics?

Computers are now better than humans at logical games and puzzles such as Sudoku, Chess, Go and so on. Mathematics can also be framed as a logical puzzle game. When will computers become better than humans at developing new mathematics and proving new theorems? Certainly this has not happened yet, but in the last few years there has been an explosion of activity, with tools such as neural networks, language models and computer theorem provers all being involved. I will survey the state of the art. Research mathematicians can currently sleep easy -- but for how long? The talk is suitable for a general scientific audience: no background in modern mathematics or computer science will be assumed.

You can find more information about speaker Kevin Buzzard here: https://profiles.imperial.ac.uk/k.buzzard