logoalt Hacker News

psychoslave • today at 5:59 PM • 1 reply • view on HN

Even exploiting bugs would be interesting and constructive, as it would reveal some things to fix.


Replies

Timwi • today at 6:42 PM

Well yeah, if we can find the bug(s). The big worry is that we might overlook the bug and go on our merry way assuming the claimed results are true when they're not. (I don't see that happening in the case of the Riemann hypothesis, but the same worry exists for all proofs that exist only in the form of Lean code, such as the massive dump from OpenAI.)