OpenAI has released three major scientific discoveries in mathematics, computer science, and physics. This follows a previous batch of math breakthroughs and marks a shift toward greater transparency and academic oversight. Source: theverge.com.
Three Major Breakthroughs
First, a proof confirms local smoothing for the Euclidean wave equation in all dimensions of two or higher. It also solves the Kakeya maximal and dimension conjectures and Stein’s spherical restriction estimates across the interior of their conjectured range.
Second, OpenAI has refuted the Exponential Time Hypothesis (ETH). A deterministic algorithm counts satisfying assignments of CNF formulas in subexponential time and polynomial space. This challenges the long-held belief that such problems require exponential time.
Third, a mathematically rigorous existence proof confirms certain interacting quantum field theories in four dimensions. This includes Quantum Chromodynamics (QCD) and a unified extension of the Standard Model. It is not experimental but provides a solid theoretical foundation for quantum physics.
External Oversight in New Releases
OpenAI now involves external experts to guide how AI-generated results are reviewed and validated. This change comes after feedback from academic panels concerned about unequal access and biased selection.
For the computer science result, a group led by Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams—original ETH proposers—will examine and explain the proof. They requested immediate public release of the AI write-up and a formal Lean proof on GitHub.
For the math result, OpenAI released the model output directly, as in its October 6 release. Researchers must build their own explanations. This approach supports academic independence and peer verification.
The AGMAI panel warns against giving early access to AI outputs. Such practices risk unequal access, borrowed reputation, and bias. They recommend AI labs publish on scholarly repositories with formal proofs, full problem transparency, and funding for research validation.
Open Letter to the Community
OpenAI released an open letter asking for community input on key issues:
- What rules should guide early access to AI-generated results?
- How should conflicts of interest be disclosed?
- How should credit be shared between AI and human researchers?
- How can students and researchers adapt to new methods while keeping critical judgment?
The letter stresses that human oversight remains essential. It calls for new norms that balance innovation with academic rigor and accountability.
Support for Researchers
OpenAI has committed $5 million to help mathematicians understand and verify the AI-generated results. Funds will support workshops, research projects, and peer review efforts.
The physics result is being guided by David Gross and colleagues—leading theorists in quantum field theory. The computer science proof is available in formal Lean format on GitHub. The math result includes model output and metadata on problem selection and training.
Scientific Caution and Limitations
Despite the claims, the scientific community remains cautious. Many mathematicians doubt the depth of understanding behind AI-generated proofs.
NYU professor Hong Wang, a Fields Medalist who solved the Kakeya conjecture in three dimensions, called the current output ‘pure insanity’. He suggests it may take years for the community to fully verify and accept the results.
OpenAI has not yet verified the findings through peer review or experimentation. The results are still under active scrutiny. The community urges independent analysis before any conclusions are drawn.
These discoveries mark a shift in how AI is used in science. They also raise deep questions about authorship, peer review, and human judgment. OpenAI’s new approach shows a move from closed development to open, collaborative validation.
Sources & further reading
Featured image: original publisher image. Via original publisher. Image source
