OpenAI unveils breakthroughs, researchers balk at gatekeeping

The announcement lands after a turbulent few months between OpenAI and mathematicians, who are still grappling with hundreds of preprints unceremoniously dumped on them at the start of October. This time OpenAI says it is working with expert "shepherds" to decide how best to share its latest breakthroughs, and each result has a different release strategy. The mathematics output was released outright, leaving researchers free to organize their own efforts to explain it; for computer science, OpenAI will assemble an external group to explain the result; physics has not been released at all, with OpenAI saying it will defer to an external group of experts.

The results themselves concern some of the most consequential unsolved problems in their fields. The mathematics claim purports to resolve the Kakeya conjecture in all dimensions — roughly, how little space is needed to rotate a needle in every direction — a problem whose three-dimensional version was among the achievements for which NYU mathematician Hong Wang won a Fields Medal earlier this year. Brown University's Brendan Hassett said the claim "looks well supported," though he had not yet checked the Lean formalization used to computationally verify results. The physics work explores quantum chromodynamics and Yang-Mills theory but does not address the unresolved "mass gap" at the heart of a Millennium Prize problem. The computer science claim, if upheld, would overturn the exponential time hypothesis, a widely held assumption closely related to P versus NP.

The frustration is less about the science than the process. AGMAI, a group OpenAI had previously pledged to consult, warned a day before the announcement that selecting certain people to receive output first is "unnecessary," borrows the reputation of selected researchers, and risks unequal access, biased selection, and delays. Xiaoyu He of Georgia Institute of Technology said the deeper issue is that fundamental results are being solved so quickly without human contribution, and that the papers are difficult to read and verify.