Mathematicians have reacted with unease after OpenAI announced that its latest AI model had solved a Millennium Prize Problem, a class of notoriously difficult mathematical puzzles that had resisted human effort for decades. The breakthrough came not from a flash of insight but from deploying vast computing resources at the problem, prompting concern about what will be left for human mathematicians and how the discipline should train future generations if AI can solve tough problems instantly.

The feat involved OpenAI running 10,000 autonomous AI agents on the problem, at an estimated cost of $15m, which carries a $1m prize. Academics including Prof Colva Roney-Dougal of St Andrews, Prof David Silvester of Manchester and Prof James Robinson of Warwick described feeling "shell-shocked" and unsettled by the pace of change, with Robinson calling it "immature playground boasting" by tech firms with huge resources. Concerns raised include researchers losing credit for problems they have worked on for months, the need to verify AI-generated proofs (echoing the 1993 discovery of a flaw in Andrew Wiles's Fermat's Last Theorem proof), and the collapse of traditional undergraduate coursework, though Prof Alexander Paseau of Oxford believes mathematics will retain its appeal.