Altman Says AI Math Rush Will Continue Despite Researchers’ Alarm
Altman Says AI Math Rush Will Continue Despite Researchers’ Alarm
OpenAI’s push into frontier mathematics first drew attention with its claim that its models had resolved a 90-year-old question tied to the Navier-Stokes equations, one of the Clay Mathematics Institute’s Millennium Problems. The company said it had identified a case in which the equations develop a singularity — a vortex spinning infinitely fast.
The breakthrough sharpened an argument already brewing among researchers. OpenAI withdrew from a Caltech “math hackathon” after current and former Caltech mathematicians warned the event was likely to harm the mathematical community. Separately, 28 Fields Medal winners said the industry’s use of difficult problems as AI benchmarks was “detrimental to the science of mathematics, and to the mathematical community.”
Their objection is not simply that machines may solve hard problems. Mathematicians have criticized OpenAI’s hurried releases for thin explanations, weak attribution and too little verification, leaving researchers to check results while hundreds more arrive. They have also objected when the company withholds details — including material on a claimed Yang-Mills result — on the advice of a selected expert group.
Altman, speaking on the Cheeky Pint podcast, acknowledged both the disruption and the emotional weight of the moment. “I think, on net, the world deserves to understand how the universe works,” he said. “But I do have empathy for the method.” He said OpenAI was trying to ensure specialized communities “get the best version of the future that they can have,” calling the task “a hard thing.”
But the CEO’s central position is firm: OpenAI should not sit on fundamental discoveries. “Humanity deserves fundamental knowledge,” he said, while conceding that the company’s mathematical sprint had briefly “get out of control.” The burst was intentional, he explained: staff fed models the hardest problems they could find through an internal channel dubbed “Model Name-Hold My Beer.”
That leaves the two sides agreeing on AI’s disruptive force but not on who should set its pace. Altman says OpenAI has more discoveries still to release; mathematicians fear the race is already outrunning the institutions meant to test and understand them.
Write a comment