In a recent Live Science article, Drew Turney reports that an AI model has solved a math problem that had gone unanswered for 80 years. According to Turney, the puzzle, called the planar unit distance problem, was first posed by Hungarian mathematician Paul Erdős in 1946, and the result is being called a milestone for what AI can do in mathematics.
Turney explains that the problem sounds simple. If you scatter dots on a flat sheet of paper, how many pairs of those dots can sit exactly one unit apart? Mathematicians had reportedly worked on it for decades, and the best result from human effort dated back to 1984.
Below are the key points from Turkey’s article:
- What the AI did. OpenAI told Live Science that one of its internal models found new arrangements of points that went further than the best earlier results. According to the company, the model got there using ideas from a different branch of mathematics, which surprised researchers because those tools were not usually linked to this kind of geometry.
- It was a general-purpose model. OpenAI said the system was not built to do math, let alone this specific problem. The company described the result as the first time an AI has solved a prominent open problem on its own, and as a sign the technology might help with frontier research.
- Humans checked the work. Outside mathematicians reviewed the proof, confirmed it, and wrote a companion paper explaining how the AI reached its answer. Thomas Bloom of the University of Manchester wrote that the original AI proof was valid but was improved by human researchers, and that people still play “a vital role” in refining such results.
- The reaction was mostly positive. Cambridge mathematician Tim Gowers wrote that if a person had submitted the paper to a top journal, he would have recommended it “without any hesitation,” adding that no earlier AI proof had come close.
- A reason for caution. Turney notes this is not OpenAI’s first big claim. Last October, staff said GPT-5 had solved several unsolved Erdős problems, but the company walked that back after experts, including Bloom, showed the problems had already been solved by people. OpenAI also stressed the technology is meant to support mathematicians, not replace them.
Whether this points to a broader role for AI in serious mathematical research is still an open question, Turney notes. For now, he says it stands as a result that human mathematicians reviewed and confirmed.
Acknowledgement: This summary was generated using AI writing tools as part of my AI Explorations process. For more, see the original article linked below.