OpenAI accused of using a mathematician's conversations to train Astra on soficity conjecture proof
iamtrask · x · 2026-09-10
- Mathematician Andreas Thom posted detailed evidence on Mastodon suggesting OpenAI may have trained Astra on conversations where he and Gábor Kun worked on Gromov's soficity conjecture — one of ten problems OpenAI announced Astra had solved.
- Quote-tweeter Valerio Capraro notes Thom is a leading sofic groups expert who spent two decades on the problem; if true this is far more than a minor attribution dispute.
- iamtrask comments it could open the floodgates of humans re-claiming credit previously given to AI.
More from Companies & People
- If Altman and Amodei both back frontier AI pacing, they should just start pacing — NathanpmYoung · 2026-09-11
- 93% of Indian healthcare leaders believe in AI, but only 11% run it in production — HealthcareLdr · 2026-09-11
- Replit courts Shopify React Native devs for its mobile push — amasad · 2026-09-11
- Netizens mock Anthropic CEO Dario Amodei's 'coders should learn to plumb toilets' remark — nevrekaraishwa2 · 2026-09-11
- Demis Hassabis' 60-minute Cambridge lecture on the future of AI — HeyAmit_ · 2026-09-11
- IIT Madras Wadhwani School of Data Science & AI opens Post-Bacc fellowship applications — ravi_iitm · 2026-09-11