Inside the Secret Meeting Where Mathematicians Struggled to Outsmart AI

A secret gathering of top mathematicians in Berkeley tested an AI chatbot, o4-mini, on complex math problems. The bot surprised researchers by solving graduate and research-level questions, often quicker than human experts. It showed human-like reasoning, tackling problems step-by-step and even solving an open question in number theory. Though the mathematicians eventually stumped the bot with 10 questions, they expressed concern about AI's rapid progress and its potential to transform mathematics, possibly relegating human mathematicians to primarily posing questions for AI to solve.