OpenAI might have stolen another major proof
Points and comments are a snapshot, not live.
Researcher Andreas Thom alleges OpenAI trained on unpublished proof work without attribution.
Valerio Capraro reports that mathematician Andreas Thom presented evidence OpenAI may have trained Astra on conversations where he and Gábor Kun worked on Gromov's soficity conjecture, one of ten problems OpenAI claimed Astra solved. Capraro, who knows Thom, says Thom spent two decades on the problem and is a leading expert. Capraro warns this could represent one of the greatest intellectual scandals in science if substantiated alongside other allegations.
What commenters are saying
Comments note the original posts are on Mastodon, not Twitter, and that the HN post links to Twitter which may require login. One user shares links to the original Mathstodon posts. Another points out that xcancel/nitter were recently disabled but notes Mastodon is open. Some express frustration that the tweet lacks a direct link to the original posts.