OpenAI has announced a significant advancement in artificial intelligence reasoning after its technology successfully addressed a mathematical conundrum that has stumped experts for 80 years.
The innovation is centered around a challenge initially introduced by Hungarian mathematician Paul Erdős in 1946, known as the planar unit distance problem. The inquiry is straightforward to grasp: if you populate a sheet of paper with dots, what is the maximum number of pairs of these dots that can be equidistant from one another? Erdős suggested that this number would increase only marginally faster than the total number of dots.
Contrary to Erdős’s assertions, OpenAI’s model made a breakthrough by leveraging various mathematical concepts to uncover a new set of arrangements that exceed Erdős’s proposed limit.
“For almost 80 years, mathematicians believed the optimal solutions were akin to square grids,” shared OpenAI on the platform X. “Our model has overturned that notion, revealing an entirely new category of structures that outperform previous beliefs.”
Though this achievement has sparked enthusiasm within the mathematical community, the larger issue remains unresolved. The AI did not present a novel rate of increase for the pairs of dots; it merely indicated that Erdős’s original limit was too conservative.
As OpenAI seeks to go public in the U.S. market, the calculations were conducted by a general-purpose reasoning model—one that breaks complex problems into digestible components—rather than a system exclusively designed for mathematical equations.
Previously, OpenAI faced challenges with Erdős-related problems, having announced a breakthrough last year based on existing literature captured by the model. However, this time, mathematicians, including Thomas Bloom—who maintains a website dedicated to Erdős’s problems—have validated OpenAI's latest findings, offering a contrast to the skepticism regarding past claims.
Bloom collaborated on a supplementary paper accompanying OpenAI’s blog post, acknowledging the AI’s ability to pursue paths that a human might consider overlooked.
Nonetheless, he emphasized that human involvement was integral to the process. “While the initial proof generated by the AI was entirely valid, it received considerable enhancements from human researchers at OpenAI and other mathematicians involved in this work. Human insight remains crucial in discussing, refining, and exploring the implications of this proof,” he explained.
Tim Gowers, another mathematician contributing to the complementary paper, characterized the accomplishment as “a significant milestone in AI mathematics.”
Andrew Rogoyski, associated with the Institute for People-Centred AI at the University of Surrey, stated that this development illustrates how AI is providing fresh perspectives on complex problems. “It’s clear that AI is influencing creative thought and will become an essential asset in future scientific endeavors,” he remarked.
