OpenAI announced on Monday that its artificial intelligence models have made several scientific discoveries. The company said the findings cover mathematics, computer science and physics.
The three discoveries relate to harmonic analysis, complexity theory and quantum field theory. OpenAI is a San Francisco-based start-up valued at $852 billion.
AI Models Tackle Math and Computer Science Problems
According to OpenAI, its models proved the local smoothing conjecture in every spatial dimension for the first time. This idea is one of the central propositions of harmonic analysis.
Harmonic analysis is a branch of mathematics. It studies how complex signals and functions can be broken down into simple waves. Before this, researchers had only proved the conjecture in three dimensions.
Alex Iosevich, a mathematics professor at the University of Rochester, said OpenAI appeared to have “cracked” the conjecture. He called it a “historic development.”
“The conjecture was formulated by my PhD adviser, Chris Sogge, so it is near and dear to my heart, and the news was a shock at first,” Iosevich told Al Jazeera.
Iosevich said he expects theoretical mathematics to “change drastically” because of AI. He said many mathematicians feel “a twinge of sadness,” but he believes that will give way to excitement about what is now possible.
OpenAI also said its models disproved the Exponential Time Hypothesis. This idea deals with how long it takes to solve complex computing problems.
The company said its models provided “mathematically rigorous” proof of quantum chromodynamics. This field studies how particles called quarks and gluons interact.
Ryan Williams, a computational complexity expert at MIT, said a confirmed refutation of the hypothesis would be a major development. “The ETH has been widely believed true since the late 1990s and a great deal of theoretical computer science was built on it being true,” he said.
Williams said the work would not have real-world uses in the short term. He said it could lead to future benefits for society.
How the Findings Are Being Released
OpenAI said it worked with three outside expert groups, which it called “shepherds.” These groups helped decide how to release the findings.
The math and computer science groups chose to make the AI-produced data public right away. The physics group decided to wait so it could write up its own explanation of the results.
OpenAI said the approach involves tradeoffs. It said picking the first groups “shapes who gets involved first and creates asymmetry of information,” and it welcomes feedback on widening participation.
The announcement follows a similar claim from rival Anthropic last month. Anthropic announced the discovery of a new enzyme that resembles CRISPR, a gene editing tool.
Scientists gave mixed reactions to Anthropic’s claim. Some called it exciting, while others said the company had overstated its importance.
For now, Williams said his team is still working to verify OpenAI’s computer science proof and determine whether it holds up.
