OpenAI posts 722 AI math preprints; mathematicians balk

OpenAI pushed 722 AI-generated math manuscripts to a public GitHub repository on October 6, grouped into 372 result families and produced by an unreleased internal model. Nature's write-up finds working mathematicians less than impressed.

The sheer number is 'staggering', wrote Alvaro Lozano-Robledo, a number theorist at the University of Connecticut, who asked 'Why is OpenAI trying to solve hundreds of math problems to be released all at once?'

Ursula Martin, a mathematician at the University of Oxford, called the drop 'tossing the community a messy first draft and expecting us to clean it up for them'. Michael Harris, a Columbia number theorist, said flatly that 'I can't bring myself to spend my time studying [these preprints], so I will not be reading these papers.'

A group letter circulating on social media put the complaint sharper: 'Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power.'

Within 24 hours OpenAI pulled three manuscripts after a sign error invalidated an argument and the constructions in two dependent papers, and revised 14 others. The batch did not include solutions to any of the five remaining Millennium Prize Problems.

Not every mathematician was horrified. Levent Alpöge, a mathematician at rival lab Anthropic, posted that 'it's obviously the most significant moment in mathematical history.' Three AI-world figures we follow shared the Nature piece within a day of posting.