Originally published on The Daily Flare.
OpenAI says its AI models have produced three new results in mathematics, computer science and physics, including a claimed refutation of a long-standing assumption about computational complexity.
Mathematics claim extends beyond three dimensions
The company announced the findings on Monday, identifying them as work in harmonic analysis, complexity theory and quantum field theory. Its mathematical result concerns the local smoothing conjecture, a proposition about how waves focus. OpenAI says the models proved it in every spatial dimension, extending beyond the three-dimensional case previously established by researchers.
The conjecture, proposed by mathematician Christopher Sogge in 1991, describes a difference between extreme and typical wave focusing: the most intense concentration should be brief, while focusing at an ordinary time is less severe. OpenAI says the result also has implications for other problems in mathematical analysis, including the Kakeya conjecture.
Alex Iosevich, a University of Rochester mathematics professor who specializes in harmonic analysis, called the claimed result a “historic development” in comments to Al Jazeera. He said the news initially surprised him, but that he expects AI to change theoretical mathematics and help researchers explore connections between abstract problems and physical questions.
The AI-produced mathematical material has been released for researchers to examine, according to OpenAI. The company describes its outputs as raw research data that requires further work to explain and place in context. Its announcement of the findings outlines a new approach involving external experts who help organize review and interpretation.
Computer science result faces independent scrutiny
OpenAI also says its models disproved the Exponential Time Hypothesis, or ETH, a widely used assumption in theoretical computer science about the difficulty of solving certain computational problems. The company’s paper describes an algorithm that counts satisfying assignments for certain Boolean formulas in subexponential time. OpenAI says the result for 3-SAT is enough to refute the hypothesis.
That claim remains under examination. Ryan Williams, a computational complexity researcher at the Massachusetts Institute of Technology, told Al Jazeera that he and other researchers were working to verify the proof. He said a confirmed refutation would be a major development because much theoretical computer science has been built on the assumption that ETH is true. The Verge reported that the review group also includes Russell Impagliazzo and Ramamohan Paturi, computer scientists associated with the hypothesis.
OpenAI made the computer science proof and its Lean formalization public at the group’s request, allowing other researchers to inspect the materials. Williams told Al Jazeera that the proposed method was unusual and might have broad implications, but said any wider benefits would be difficult to predict and were not expected in the short term. The central question for now is whether independent specialists can confirm that the argument is valid.
Physics materials held for outside interpretation
The third result concerns quantum field theory, the mathematical framework used to describe fundamental particles and forces. OpenAI says its models produced a mathematically rigorous existence proof for theories involving interacting fields, including quantum chromodynamics, which describes interactions between quarks and gluons. The Verge reported that the company also characterized the work as including a unified completion of the Standard Model.
Unlike the math and computer science materials, the physics artifacts have not been released publicly. OpenAI said the outside group advising on the work chose to delay publication while its members prepare their own interpretation. The Verge reported that the group is led by physicist David Gross and includes Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov.
The scope of the claim matters. Scientific American reported that the work addresses the existence of certain quantum field theories, while the separate question of whether the Yang-Mills theory has a mass gap remains open. The distinction means the announcement should not be read as a resolution of the entire Yang-Mills problem.
OpenAI says its release process was shaped by consultations with academic groups and that its collaborators participate voluntarily, without being asked to sign nondisclosure agreements. The company also acknowledged a drawback: choosing initial experts can affect who gets access first and create unequal information. The Advisory Group on Mathematics and Artificial Intelligence had urged AI companies not to designate selected researchers to shepherd results, according to The Verge, underscoring continuing debate over how claims generated by private AI systems should be shared, checked and credited.
Top comments (0)