OpenAI floods GitHub with 700+ AI-written math preprints

OpenAI posted more than 700 AI-generated mathematics preprints to GitHub on October 6, written by an unreleased internal model that had been pointed at roughly 4,000 problems and claiming progress on 372 of them. Nature reports that none of the 722 manuscripts offered a solution to any of the Clay Mathematics Institute's five remaining Millennium Prize Problems.

Reactions landed hard. 'I will not be reading these papers,' said Michael Harris, a number theorist at Columbia University. Alvaro Lozano-Robledo, a number theorist at the University of Connecticut in Storrs, called the simultaneous release 'staggering' and asked, 'Why is OpenAI trying to solve hundreds of math problems to be released all at once?' Oxford mathematician Ursula Martin described the strategy as 'tossing the community a messy first draft and expecting us to clean it up for them.'

Not everyone recoiled. At rival lab Anthropic, mathematician Levent Alpöge called the drop 'obviously the most significant moment in mathematical history', while allowing there would be 'sad stories' for researchers whose projects were preempted.

Three manuscripts were pulled a day later after a sign error invalidated one paper's argument and the constructions in two dependents. OpenAI said its goal is to 'provide tools for mathematicians and scientists' and that it evaluates its models 'against difficult frontier problems in all sciences'. The story was moving fast through the researcher circles on our radar: three of the people we track shared it the same day.