OpenAI Announces Major Breakthroughs in Mathematics, Computer Science, and Physics, but Researchers Feel Fatigued by Its Information Disclosure Strategy

Significant Breakthroughs across Three Core Scientific Fields

The newly announced results tackle some of the most consequential open problems in their respective areas. The mathematics result reportedly resolves the Kakeya conjecture in multiple dimensions, a problem that concerns the minimal space required to rotate a needle or pencil in all directions. This problem's three-dimensional version was among the work recognized by the Fields Medal awarded to NYU's Hong Wang. Brown University's Brendan Hassett commented that OpenAI's claims seem well supported, though he has not yet examined the Lean formal verification used. In physics, the work explores quantum chromodynamics within the Standard Model, analyzing its relation to Yang-Mills theory, one of the Millennium Prize problems, though not tackling the unresolved mass gap issue. In computer science, OpenAI claims to disprove the exponential time hypothesis, which suggests certain problems cannot be solved significantly faster than brute force, closely connected to the Millennium Prize

P versus NP

problem.

Mixed Responses on OpenAI's Release Strategy

OpenAI is implementing differentiated release methods guided by appointed expert 'shepherds.' The mathematics findings are published as AI outputs, allowing researchers free rein to explore and explain them. For the computer science development, an external group of researchers will be assembled to explain the results. The physics findings remain unreleased pending advice from outside experts. Despite these efforts, some in the academic community criticize OpenAI's handpicking of experts to review results as limiting transparency and overall scholarly access. Andreas Karch, a physics professor at the University of Texas at Austin, labeled the selection process a 'disaster' and suggested simultaneous public release alongside expert review to democratize access. The Advisory Group on Mathematics and Artificial Intelligence (AGMAI) had earlier cautioned against pre-selecting reviewers, warning that it risks inequity and delays, and recommended releasing work openly on scholarly repositories supplemented by funding for further analysis.