@highergeometer@mathstodon.xyz
Post #4292320
2026-07-27 06:26 UTC
@gregeganSF@mathstodon.xyz Nice to hear at least one definitive negative result from people trying to use an AI tool to solve a famous open problem (the idea was in the air, AI couldn't get it over an extended period of time, then the human team solved it)
Replies (1)
-
@gjm@mathstodon.xyz 2026-07-27 11:39
@highergeometer@mathstodon.xyz @gregeganSF@mathstodon.xyz At present the AIs seem to be much better, relative to humans, at finding counterexamples to conjectures than at proving them. Cf. Kevin Buzzard's "Human mathematicians are being outcounterexampled". Of course there's no guarantee that that state of affairs will last.