OpenAI publishes 719 AI-written math manuscripts and shakes up the field

In early October, OpenAI published a collection of 719 AI-generated mathematical manuscripts, grouped into 372 results. Among them is a proof that certain functions linked to the Riemann hypothesis have no zeros beyond a bound, which it calls the quasi-Riemann hypothesis.

The mathematician Terence Tao, a Fields medallist, told Numberphile that it marks the end of the age of heroes in mathematics: proofs are now plentiful, but less is understood. He has read about ten of the papers: some have brilliant ideas, others are incomprehensible and a few have errors. He criticised OpenAI for consulting no one, and the race towards artificial general intelligence for sacrificing understanding and safety. Steven Strogatz spoke on The Daily about his profession's grief.