logoalt Hacker News

akomtu • today at 6:05 PM • 1 reply • view on HN

This kind of proofs don't help humanity at all. AI uses a gobbledygook language that's incomprehensible by humans and the only outcome of its proof is a 0/1 result that you either believe blindly or not, for nobody can really verify it.


Replies

ducttapecrown • today at 6:17 PM

This is not true. In the processes I've read about, AI produces natural language proofs first, which it then translates into Lean code which is verified by the Lean kernel.

➕ show 1 reply