Join Nostr
2025-12-30 17:14:34 UTC
in reply to

Bartosz Milewski on Nostr: nprofile1q…8kkj3 I'm pretty sure that we can train the AI to optimize the size of ...


I'm pretty sure that we can train the AI to optimize the size of the proofs. As for being able to understand and interpret the results, that's a different story. Only a handful of mathematicians are able to understand the Wiles's proof of the Fermat's theorem. An AI generated proof could be beyond the scope of understanding of any human mathematician. Would we reject it?