‘Mozart of math’ Terence Tao says AI race to solve math’s hardest problems is harming mathematics

The University of California, Los Angeles professor, nicknamed the "Mozart of math," issued the warning after OpenAI announced last week that its latest AI model had solved the Navier-Stokes problem, one of seven Millennium Prize Problems. The Clay Mathematics Institute established the problems in 2000, offering a $1 million prize for solving each one.

Tao, who won the Fields Medal, mathematics' highest honor, in 2006, said competition to make discoveries first has always existed in the field, but its difficulty previously acted as a natural constraint. "Now, there's no speed limit, and suddenly things are breaking down," he told British science magazine New Scientist.

His concern is not that AI can solve difficult problems, but that tech companies are racing to produce results without doing the slower work needed to integrate those discoveries into mathematics. He said mathematical discoveries must pass through five stages to become useful: they must be created, checked, explained, accepted and eventually incorporated into the knowledge taught to future generations.

AI companies, he said, are largely concentrating on the first two. "These companies are dumping carcasses of raw meat onto our communal village table and saying: 'Here you go, I solved your food problem.' And then they just leave."

Terence Tao, a professor of mathematics at the University of California, Los Angeles and Fields Medal winner. Photo courtesy of UCLA

He said the race rewards being first rather than producing work that is accurate, clear and useful to other mathematicians.

"This is not how science is supposed to work," he said, arguing that AI companies focus on prestigious discoveries while leaving mathematicians to explain, communicate, teach and review the results. "All that work is left to us to clean up, and it's demoralising."

The Navier-Stokes equations describe how fluids move and are used to model phenomena ranging from air and water flows to weather. OpenAI deployed 10,000 AI agents, autonomous systems capable of carrying out tasks, to tackle the problem at an estimated cost of $15 million, according to The Guardian.

Before OpenAI announced its solution, Tristan Buckmaster of New York University and Levent Alpöge of Anthropic had been conducting related research using OpenAI products. They later suspected the company's model had learned from their unfinished work, which OpenAI denied after an investigation.

Tao said the episode could make mathematicians less willing to share unfinished work, as AI makes information about what they are studying increasingly valuable to competitors.

Other mathematicians have voiced similar concerns. James Robinson of the University of Warwick criticized tech companies for using major unsolved problems to demonstrate their AI capabilities, describing the competition as "immature playground boasting."

Researchers have also raised concerns that AI could reshape mathematical careers and education as systems become capable of solving problems that once required months or years of human work.

Tao nevertheless expects mathematicians to eventually incorporate AI discoveries into the field.

"It will eventually end up in textbooks, but it's just a completely sordid way to get there," he said.