Insider Brief
- OpenAI is opening three AI-generated research claims in mathematics, computer science and physics to outside examination through different academic review and release approaches.
- The claims concern the local smoothing conjecture, the Exponential Time Hypothesis and mathematical existence proofs for quantum field theories, including quantum chromodynamics.
- Outside researchers will assess the claims and their implications, while practical applications would require further work.
OpenAI is bringing AI-generated research claims in mathematics, computer science and physics to outside scientists, testing different approaches to checking and explaining the results.
The announcement pairs three claimed advances with different release strategies: public examination of mathematics materials, an organized collaboration around computing research, and a physics group preparing its own account before releasing the underlying work.
For AI research, the efforts address a question beyond whether a model can produce a convincing answer: How can scientists establish that its reasoning supports a new discovery?
Mathematical verification software can help check a proof, but researchers also need to understand what it establishes, how it relates to earlier work and whether it offers methods others can use.
A Claim About Computing’s Limits
OpenAI says its models disproved the Exponential Time Hypothesis. Russell Impagliazzo and Ramamohan Paturi of the University of California, San Diego, and Ryan Williams of MIT will lead examination of the result.
The hypothesis concerns the difficulty of solving certain logic problems. A computer must determine whether a collection of true-or-false choices can satisfy all the conditions it receives.
Checking a proposed answer can be straightforward. Finding an answer can require searching through possibilities that multiply rapidly as the problem grows.
The hypothesis holds that even the best algorithm cannot avoid exponential growth in the time required to solve these problems in the worst case. Research by Impagliazzo, Paturi and Francis Zane connects that assumption to other tasks, including graph coloring and identifying particular groups within networks.
A disproof would challenge assumed limits on possible algorithms. It would not automatically make large calculations practical: A method can improve substantially on exponential growth and still take too long to be useful.
Mathematics of Waves
The mathematics claim concerns the local smoothing conjecture, which OpenAI says its models proved in every dimension, with consequences for the Kakeya and restriction conjectures.
Local smoothing concerns how waves behave over time. A wave can become sharply concentrated at a particular moment; the mathematical question asks how much that concentration is reduced when its behavior is considered across a period of time.
The problem belongs to harmonic analysis, a field that studies functions by examining their oscillations and frequency components. Earlier work by Larry Guth, Hong Wang and Ruixiang Zhang established a sharp estimate connected to local smoothing for waves in two spatial dimensions. Their published paper provides part of the research background against which a broader claim must be assessed.
Foundations of Particle Physics
The physics effort concerns claimed mathematical existence proofs for quantum field theories, including quantum chromodynamics. A group led by David Gross will examine the work.
Quantum field theory describes particles and their interactions through fields governed by quantum mechanics. Quantum chromodynamics, or QCD, describes the strong interaction involving quarks and gluons.
A mathematical existence proof asks whether a theory can be constructed with the required properties on a rigorous foundation. This is a different task from showing that its predictions agree with experiments. Physicists can obtain successful predictions from a theory while unresolved questions remain about its mathematical construction.
The related Yang-Mills Millennium Prize problem requires both establishing the theory’s existence and proving a mass gap—a minimum energy separating the vacuum from its excitations. An existence proof alone would address only part of that problem.
Checking and Understanding
OpenAI says its mathematics and computing materials include proofs checked in Lean, software for verifying mathematical reasoning.
Lean requires proofs to be expressed in a precise mathematical language, then checks whether the reasoning follows the system’s logical rules. That provides stronger evidence than a written explanation that merely appears persuasive.
Researchers must still establish that the formal statement captures the intended problem and explain what follows from it. For the three efforts, that work will determine how AI-produced proofs become results the wider scientific community can assess and build upon.
