OpenAI model's PDE proof compiles in Lean; researcher likens fuss to whining over Perelman
RexDouglass · x · 2026-09-15
Mathematician Scott Armstrong pushes back on criticism of OpenAI's model-generated PDE proof: yes, the paper isn't polished, but he finds the outrage overblown. The problem had stumped the field for 100 years, with most top PDE analysts having tried and failed — and OpenAI's model produced a proof that actually compiles in Lean.
He argues that's insanely valuable even if a team needs a year to rewrite it readably (he expects far less), citing Perelman's Poincaré proof: a bare sketch posted to arXiv that took the community 4-5 years to decode, with nobody claiming it "set back differential geometry."
Related event: OpenAI's Navier-Stokes Claim Draws Both Praise and Skepticism(7 posts)→
More from Fun
- 'Mom, shhh, I'm talking to Devin AI': the meme every dev relates to — marvinvonhagen · 2026-09-15
- 'A simulated Uber Eats delivery doesn't make anyone full' — the John Searle AI meme — dioscuri · 2026-09-15
- AI psychosis is self-reinforcing, and benhylak says those jokes are cries for reassurance — teodorio · 2026-09-15
- Cassini's Saturn mission ended 9 years ago: planned for 3, lasted 13 — AIFlow_ML · 2026-09-15
- "Future computers only run ChatGPT" meme meets the token-spewing punchline — teodorio · 2026-09-15
- Buy a domain, get a cert, sell it: lcamtuf flags a free MITM path — OwariDa · 2026-09-15