Working with academic communities to build scientific understanding

OpenAI’s announcement describes AI-assisted results in harmonic analysis, complexity theory and quantum field theory, including a claimed disproof of the Exponential Time Hypothesis, a result on the Kakeya conjecture, and work on the mathematical existence of QCD. Commenters expressed astonishment at the ETH result, noting that the hypothesis underpins many complexity-theory arguments. Some wondered whether it shifts expectations about P versus NP; others emphasized that a subexponential algorithm does not itself establish P=NP. Several saw the breakthroughs as a striking sign of what AI might make possible, while others questioned whether similar progress would extend to less formally verifiable fields such as biology. One commenter welcomed the results as potentially useful in applied physics.

Views diverged over how the work should be shared and scrutinized. Some praised the plan to publish the computer-science proof and Lean formalization, involve academic experts, and fund independent examination; they saw this as an improvement on earlier releases. Others suspected a publicity strategy, objected to AI-generated work burdening researchers with verification, or considered the proposed funding inadequate relative to the costs imposed on science. The decision to withhold the physics artifacts while researchers prepare them drew particular curiosity. Commenters also debated whether the results represent a new, more empirical mode of mathematics and computer science, and whether AI’s progress will remain concentrated in areas where proof-checking provides a clear signal.