Fields Medalists Warn AI Labs Are Hurting Real Mathematics

Some of the world’s most respected mathematicians are warning that AI companies may be changing mathematics in the wrong way.

Terence Tao and 24 other Fields Medalists have signed a declaration arguing that the race by AI companies to solve major mathematical problems as benchmarks is damaging to the discipline. The statement, published on Math and AI and reposted by Tao on his personal blog, says the goals of AI companies and the goals of the mathematics community are now severely misaligned.

Their argument is subtle, and it is not simply that mathematicians are angry because AI is getting better at math. The concern is that mathematics is not only about final answers. It is about concepts, methods, teaching, collaboration, attribution and the slow process through which a result becomes part of human understanding. A famous unsolved problem is not just a trophy. It is often a guidepost that helps researchers build new ideas around it.

The declaration says AI labs are treating hard math problems like public scoreboards. If an AI system can solve a famous problem first, that becomes a marketing event and a capability milestone. But the mathematicians argue that rushing to announce solutions can leave little room for proper write-ups, citation of previous work, isolation of new methods and the patient explanation needed for the community to absorb the result.

This comes after OpenAI said an AI system had solved a major math prize problem, triggering questions from mathematicians about the process, credit and verification. TechBooky previously wrote about that dispute and why the mathematician’s side of the story matters. The new declaration broadens the issue from one result to the future of intellectual work itself.

The biggest warning here may apply far beyond mathematics. If AI systems can quickly produce outputs in law, software, science, journalism, design or engineering, society will have to decide what happens to the training process that used to create experts. A correct answer can be useful, but a field also needs people who know why the answer is correct, how it connects to older ideas and what question should come next.

That does not mean AI should be kept out of mathematics. The signatories say AI could enhance and accelerate genuine mathematical study. The issue is control and purpose. If AI is used to help humans understand more deeply, it can be powerful. If it is used mainly by companies to win attention and prove model dominance, the field may end up with more answers and less understanding.

For the AI industry, this is another reminder that capability is not the same as wisdom. Solving a hard problem may impress investors. Building a healthier relationship with the communities whose work made those systems possible will require more humility than a leaderboard.