OpenAI Unveils Three Major Scientific Breakthroughs in Math, CS, and Physics with New Academic Collaboration Model

In a significant development for the intersection of artificial intelligence and fundamental science, OpenAI has announced three major discoveries spanning mathematics, computer science, and physics. This announcement comes less than a month after the company released hundreds of manuscripts claiming solutions to long-standing mathematical problems. The key distinction this time lies in the methodology of dissemination; OpenAI has adopted a new advisory framework involving external experts, whom it describes as "shepherds," to guide the validation and explanation of the results in computer science and physics. For the mathematical breakthrough, the company followed advice to release the raw AI output directly, allowing the global research community to develop their own explanations and verifications. The most controversial and potentially paradigm-shifting result pertains to computer science, where OpenAI claims to have proven that the Exponential Time Hypothesis (ETH) is false. The ETH, widely believed to be true since its proposal in 1999, posits that certain computational problems require exponential time to solve. The new paper presents a deterministic algorithm that exactly counts satisfying assignments of fixed-width CNF formulas in subexponential time and polynomial space. Specifically, the case of 3-SAT refutes the hypothesis. To ensure rigorous scrutiny, an independent working group led by prominent computer scientists Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams has been tasked with examining and explaining the proof. At the group’s request, the materials, including the AI-generated write-up and Lean-formalized proof, were made public immediately on GitHub. In the realm of physics, OpenAI has announced a mathematically rigorous existence proof for certain interacting quantum field theories containing gauge fields and fermions in four dimensions. This includes Quantum Chromodynamics (QCD) and a unified completion of the Standard Model of particle physics. While the full details have not yet been released, the validation process is being led by David Gross, a Nobel laureate, alongside a distinguished group including Michael Douglas, Arthur Jaffe, Igor Klebanov, and Nikita Nekrasov. This high-level oversight aims to lend credibility to the complex theoretical physics claims generated by the AI model. The mathematical discovery involves a proof of the local smoothing conjecture in every spatial dimension at least two. This proof utilizes an auxiliary partial differential equation to establish Sogge’s local smoothing for waves. Furthermore, it yields the Kakeya maximal and dimension conjectures and Stein’s spherical restriction estimates throughout the interior of the conjectured range. Notably, just a few months ago, NYU professor Hong Wang was awarded the prestigious Fields Medal for resolving the Kakeya conjecture in three dimensions, highlighting the intense current focus on these specific mathematical problems. OpenAI acknowledges that this new release strategy is not without challenges. A panel of mathematical advisers, including those from the AGMAI initiative, recently recommended that AI companies avoid designating specific individuals for early access to outputs, citing issues such as unequal access and biased selection. Instead, they advocated for releasing outputs on scholarly repositories with formalized proofs and transparent information on problem selection. In response, OpenAI has published an open letter seeking community input on broader ethical and methodological questions, including how to balance credit between discovery and verification, and how to support researchers as AI methods evolve. This transparent approach marks a maturation in how AI-generated scientific breakthroughs are handled, aiming to foster trust and rigorous peer review in an era of rapid AI-driven discovery.