OpenAI Presents Scientific Proofs Challenging Long-Standing Computer Science Hypotheses

OpenAI announced on Monday that its systems have successfully produced mathematical arguments challenging several major problems that have resisted researchers for decades. The computer science claim centers on the Exponential Time Hypothesis, which posits that specific Boolean satisfiability problems cannot be solved in subexponential time. This foundational assumption underpins numerous conditional results regarding the computational difficulty of tasks across graph theory, databases, computational biology, and geometry.

According to the company, its model generated a natural-language disproof on October 16, followed by a formalization completed using Lean software on October 17. Lean verifies whether a mathematical argument adheres to specified formal rules, although specialists must still examine the underlying definitions, assumptions, and overall significance of the results. Russell Impagliazzo and Ramamohan Paturi of the University of California, San Diego, alongside Ryan Williams of the Massachusetts Institute of Technology, will lead an external examination group to scrutinize the claim.

The review group has requested all raw materials to facilitate a thorough independent assessment, and OpenAI has agreed to finance a conference dedicated to exploring the broader implications. If the result is upheld, researchers will be forced to reassess numerous conclusions whose complexity guarantees depend on the hypothesis. However, the company noted that the theoretical result does not independently settle the separate P versus NP question or immediately yield improvements in commercial computing.

In addition to the computer science claim, the announcement covers advancements in mathematics and physics. The mathematical development involves the local smoothing conjecture proposed by Christopher Sogge in 1991, which describes how extreme concentrations of waves diminish when averaged over time. OpenAI claims its system proved this conjecture across all relevant dimensions, utilizing a new auxiliary partial differential equation that was formalized in Lean shortly after its generation.

The physics claims involve rigorous mathematical constructions of interacting quantum field theories in four dimensions. These include existence results for quantum chromodynamics, which describes the strong nuclear interaction, a unified completion of the Standard Model of particle physics, and a potential existence proof for pure Yang-Mills theory. Nobel laureate David Gross will lead a separate physics review group alongside Michael Douglas, Arthur Jaffe, Igor Klebanov, and Nikita Nekrasov to evaluate these findings.

OpenAI emphasized that its collaborators are participating voluntarily without financial compensation, are unbound by nondisclosure agreements, and retain the freedom to publish independent conclusions. The organization stated it will not claim academic authorship for discoveries generated by its internal models, while committing to supporting broader independent scientific examination through funding administered by the Simons Foundation.