OpenAI’s Math Blitz Stuns Researchers—and Leaves Them Checking the Fine Print

OpenAI’s Math Blitz Stuns Researchers—and Leaves Them Checking the Fine Print
The turbulence began in September, when OpenAI said an internal model had apparently solved Navier-Stokes, one of mathematics’ Millennium Prize Problems. The claim electrified the field—and drew accusations that the company’s systems might have benefited from researchers’ unfinished work, which OpenAI denied.

In the aftermath, the Institute for Advanced Study-hosted Advisory Group on Mathematics and Artificial Intelligence urged labs to publish results promptly through established academic channels, disclose the model, prompts and computing costs, and avoid turning mathematical releases into product marketing.

Then came Tuesday’s deluge: 722 manuscripts arranged into 372 families of findings, spanning hundreds of open questions. OpenAI described the work as produced by an internal frontier model; Greg Brockman framed the ambition as moving “towards acceleration of scientific discovery and improving quality of life for everyone.” For admirers, the sheer scale signaled a sharp expansion of what AI can do. Oxford mathematician Martin Bridson called the release “breathtaking,” while Rutgers department chair Alex Kontorovich said one proof would merit “an instant Fields Medal” if a human had produced it.

But the same volume made scrutiny harder, not easier. Researchers told The Verge that nearly 400 AI-generated results across more than 700 manuscripts could take years simply to absorb—before anyone can establish which proofs are sound, original or useful. Stephen Wolfram captured the skeptical view: “You can make a trillion theorems easily. The problem is most of those theorems are not ones that anybody will care about.”

The caution was not theoretical. By Thursday, OpenAI had withdrawn solutions to at least three problems after apparent errors. Critics including the Association for Human Mathematics called the mass publication “not a demonstration of scholarship, but a demonstration of power,” while supporters such as Toronto mathematician Dan Litt argued new answers remain valuable—provided society continues to fund and cultivate human expertise.

That divide now defines the release. Formal verification may increasingly automate proof-checking—Yann LeCun says the emphasis can shift toward new concepts and abstractions—but mathematicians insist the field cannot become a permanent cleanup crew for an AI lab’s output.

https://foxvector.com