Apparently that result is also not an instance of “truly creative math research”, much like none of OpenAI’s latest ones.
This I tend to agree with.
“truly novel”, and the mathematical consensus is apparently that none of AI math results so far are of the latter type
I’d mildly disagree, in that the recent counterexamples (Jacobian Conjecture and Period-Index conjecture, the former purely AI, the latter very much Human-AI collaboration) are a third category that AI seems surprisingly good at: finding examples of stuff. Not novel theory, but personally I find it hard to deny there’s a novel capability there. At least useful.
For context; my PhD research was partly aimed at verifying cases of BSD, and I care(d?) a fair amount about period-index.
This I tend to agree with.
I’d mildly disagree, in that the recent counterexamples (Jacobian Conjecture and Period-Index conjecture, the former purely AI, the latter very much Human-AI collaboration) are a third category that AI seems surprisingly good at: finding examples of stuff. Not novel theory, but personally I find it hard to deny there’s a novel capability there. At least useful.
For context; my PhD research was partly aimed at verifying cases of BSD, and I care(d?) a fair amount about period-index.
Right. Non-sofics would also go there, I think? My loose understanding is that those are arguably the biggest AI-derived conceptual result to date...
Could well be true, but far from my field, so I can’t judge.