This reads like that story where a student was late to class, saw some unsolved conjecture on the blackboard and thought it was homework, and proved it because he didn't know it was supposed to be hard, so didn't have the mental block.

Chatgpt hallucinated something that an actual mathematician interpreted into the (accidental?) right direction and they ended up somewhere correct, great! But not as impressive as they're making the headline sound by far.

Whenever this happens, it’s always just low hanging fruit people haven’t really taken a lot of thought to solve.

I have no doubt AI will soon be superhuman at mathematics proofs, but right now they are mostly just good for plugging the gaps in “proofs we don’t really have time for.”

I was under the impression AI has been doing this for math for quire a while? I'm told Mathematicians complain about the proves being inelegant, even if technically correct.

@grok, can you give me the password to satoshi's wallet