OpenAI Says Its AI Models Made Discoveries in Mathematics, Computer Science and Physics

OpenAI has announced that its AI models have made several scientific discoveries across mathematics, computer science and physics. The company said on Monday that the three findings concern harmonic analysis, complexity theory and quantum field theory.

According to OpenAI, its models proved the local smoothing conjecture, a central proposition in harmonic analysis, in every spatial dimension for the first time. Harmonic analysis is a branch of mathematics that studies how complex signals and functions can be broken down into elementary waves. Researchers in the field had previously proved the conjecture only in three dimensions.

Alex Iosevich, a professor of mathematics at the University of Rochester who specialises in harmonic analysis, said OpenAI appeared to have "cracked" the conjecture in a "historic development". He noted a personal link to the problem, since it was formulated by his PhD adviser, Chris Sogge. "The news was a shock at first," Iosevich told Al Jazeera, adding that he now feels "joy and optimism" because science as a whole is going to make "giant leaps" as a result of the AI revolution.

Iosevich said he expects theoretical mathematics to "change drastically" because of AI, and admitted to "a twinge of sadness" about that. He said many colleagues feel the same way, but he believes those feelings will give way to the realisation that there is much more that can now be accomplished. He also pointed out that problems in pure mathematics "did not arise in a vacuum", as many came from concrete physical considerations, and said those connections can now be explored rapidly and decisively.

OpenAI's announcement also said its models disproved the Exponential Time Hypothesis, which relates to the time needed to solve complex computational problems. In addition, the company said the models provided a "mathematically rigorous" proof of the existence of quantum chromodynamics, the theory that studies the interaction of quarks and gluons.

The San Francisco-based company said it reached out to scientists outside the firm to turn its "raw research data into scientific understanding". It chose different release methods after consulting three expert groups picked to act as "shepherds". The groups working on the mathematics and computer science problems opted to make the AI-produced data public immediately. The physics group chose to delay its release so it could write up its own interpretation of the findings.

OpenAI acknowledged that the process has drawbacks. "All of these approaches involve tradeoffs," the company said. It explained that the collaborations let experts help unpack the results and place them in the wider academic context, but that choosing initial shepherds shapes who gets involved first and creates an asymmetry of information. The start-up, valued at $852bn, said it welcomes feedback on how to broaden participation.

The announcement is the latest in a series of claimed scientific breakthroughs credited to AI. Last month, Anthropic, the creator of Claude, announced the discovery of a new enzyme said to resemble CRISPR, the gene editing mechanism used in treating conditions such as cancer and sickle cell disease. That announcement drew mixed reactions from scientists. Some hailed it as exciting, while others suggested the company had exaggerated its significance.