25 Fields Medalists declare war on the benchmark race — and Google just bought the other side of it

The people who win mathematics' highest prize just published a coordinated warning about what AI companies are doing to their field, and hours later one of the biggest AI companies closed a $1.5 billion-plus deal to accelerate exactly what they warned against.

Terence Tao and 24 fellow Fields Medalists published a joint declaration today arguing that AI companies' push to solve mathematical problems as a benchmark is "detrimental to the science of mathematics" — and that the goals of AI companies and the mathematical community are "severely misaligned." Every initial signatory holds a Fields Medal, the discipline's highest honor: Tao, Manjul Bhargava, Pierre Deligne, Peter Scholze, James Maynard, Maryna Viazovska, Caucher Birkar, Elon Lindenstrauss and 17 others, spanning winners from 1978 through this year's medalists.

The declaration's argument is more specific than a generic "AI bad" statement, and it's worth reading closely. Famous open problems, the signatories write, have served as "landmarks and lighthouses" — the point of solving one was never the true/false answer but the new ideas and methods that got unlocked, then transmitted through talks, discussion, simplification and eventually textbook presentation. That transmission chain is the danger zone: the declaration warns that "the mass production at faster and faster pace of 'true/false' statements could destroy fertile ground instead of breathing life into new ideas." It also flags attribution and plagiarism problems — solutions "announced in a rush, leaving no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others."

The signatories frame this as a template for every intellectual profession, not a parochial math grievance: "We are witnessing a general threat to intellectual work, with misalignment between the outcome of the use of AI and its initial purpose." Fields medals were never the target here — the letter is explicit that the questions math is facing now are questions "all of humanity might face."

The urgency is real, and it's grounded in what's actually been happening on Tao's own blog. This week alone, posts appeared announcing what look like LLM-assisted results on the Hodge conjecture, the existence of non-sofic groups, and stable singularities of the Euler equations — landmark-level targets landing at a pace no human community could absorb. That's precisely the "mass production of true/false statements" the declaration describes, arriving while mathematicians are still arguing about how a counterexample to the Jacobian conjecture was even produced. And notably, the mathematicians chose speed over process: Tao writes that they skipped the lengthy consultative process used for the earlier Leiden declaration because "the urgency of the situation was such that we needed to release a statement sooner rather than later." Like Leiden, they're inviting further signatures — so this list will likely grow past 25.

Hours after the declaration went up, Business Insider reported that Google appears to have completed its talent deal with AI coding startup Mechanize, reportedly for $1.5 billion or more — with departing staff updating their LinkedIn profiles as evidence. Google had reportedly been circling the startup, which builds agents aimed at automating knowledge work, since at least August. The juxtaposition writes its own punchline: on the same day, the top of mathematics formally asked AI companies to stop optimizing for benchmark victories, one of those companies spent ten figures acquiring the team building the machinery to win more of them. Whatever the mathematicians are asking for, the market is bidding the other way.

What to watch: whether major non-Fields signatories pile on, and whether any AI lab actually changes how it announces math results — proper writeups, credited prior work — rather than just issuing statement-of-concern acknowledgments.

If the people who understand these proofs best say the benchmark race is damaging the field, what exactly would a "responsible" announcement of an AI-solved Millennium Problem even look like?

Sources: Terence Tao — What's new · Math and AI declaration · The Economist · Business Insider · Techmeme