We asked [mathematicans about an incomprehensible AI proof of GRH] before it happened (13min38s)
12 points by dzwdz
12 points by dzwdz
Over the last year and a half, we've interviewed mathematematicians about AI in math.
"How would you feel if AI produced a Lean-verified proof of the Riemann hypothesis, incomprehensible to humans?" was one common question.
Here's what they said.
Posting to be merged into https://lobste.rs/s/cmifpg/levent_alpoge_claims_proof_riemann, I think this is good commentary on the topic (includes thoughts from e.g. Lean's creator).
I refrained from posting this because I thought this is off-topic (and there isn't a fitting "ai" tag you could use to filter this out), but since the original story seems to have been well-received...
Interviewees, in order (errors mine):
Would you trust a Lean-verification of a huge AI-generated proof of RH, or could it be exploited? Would such a proof be good for math?
How plausible is it that, [as AI drastically improves], we will [be surprised] at how little insight we got from some Lean solution proving the Riemann hypothesis or something? Or [is] a necessary condition of solving the Riemann hypothesis that the constructions which are made, the definitions which are created, even in the Lean program, have to advance our understanding of mathematics?
Sorry for the few X links in there, they're the best websites I could find. 3blue1brown has posted transcripts of some videos in the past, so I really hope we'll get a transcript of this one too, as this is some fantastic journalism.
Many thanks to Daria Ivanova for organizing and conducting most of these interviews, and kicking off the whole project.