OpenAI's AI solves decade-old math problems, and mathematicians debate what it means for the profession
A big AI milestone in math that is exciting for research but unsettling for the people who do it.
- OpenAI's system cracked 10 long-unsolved math conjectures, results a human would have been rewarded for.
- The problems are real but niche — strong work, not the top breakthroughs of the year, and mostly solvable with existing methods.
- AI shines where it can brute-force millions of tries or where answers are easy to check by computer, less so on deep original theory.
- The proofs were verified rigorously in a system called Lean, but most AI-generated math still needs humans to confirm it is correct.
- Some mathematicians feel real dread that machines are taking over the discovery they love, even as others predict a surge in demand for mathematicians to check and build on the flood of new work.
Outlook: These tools are expected to keep improving fast, forcing changes in how math is taught and how the profession is structured.