That Proof Was Downstream of Mathematicians' Chats — but Is 'Direct Training' Even Real?

shaurizard · x · 2026-09-10

Responding in the thread about labs using frontier mathematicians' chats, shaurizard concedes the data was valuable — as he understands it, the recent proof was downstream of those earlier discussions — but questions the 'direct training' framing: he believes the information reached the model, not the way people describe it. He doesn't object to the tactics since the data was given voluntarily, but argues anyone with real objections should protest the right approach.

Related event: OpenAI Claims Navier-Stokes Breakthrough, Sparking Authorship Dispute(465 posts)→

Original post →

More from AGI Musings

AGI Musings channel →