color: var(color-red-500)

by

An AI Cracks an 80-Year-Old Geometry Puzzle. What Do Mathematicians Make of It?

In 1946, mathematician Paul Erdős presented a challenge called the unit distance problem. Now, after 80 years, an artificial intelligence has found solutions that surpass previously believed limits. The AI has demonstrated the existence of at least n^(1+δ) pairs of points that are a unit distance apart (where δ is a small positive number). Mathematicians at Princeton, including Tim Gowers and Arul Shankar, have confirmed this result, calling it a major breakthrough.

  • Key Takeaways:

    • OpenAI solved Paul Erdős’ 1946 puzzle with n^(1+δ) unit-distance constructions.
    • Princeton verified the result, giving AI a 2026 credibility boost in mathematics.
    • Tim Gowers says the advance could influence cryptography and proofs beyond geometry.

    • An 80-year-old geometry riddle finally budged when an OpenAI system stitched together an unlikely construction that beat long-standing expectations. The unit distance problem, posed by Paul Erdős in 1946, asks how many pairs of points exactly one unit apart can exist among n points in the plane; the AI found configurations that grow faster than the classic playbook allowed. Princeton mathematicians checked the work, and heavyweights like Tim Gowers and Arul Shankar took notice. Beyond bragging rights, the result hints at a new kind of collaborator for math, one that uses general inference to push past human heuristics.

    AI cracks 80-year-old mathematical mystery with breakthrough solution

    Certain challenges persistently test human endurance. One such problem, introduced by Paul Erdős in 1946, is surprisingly simple to state: given a set of points on a flat surface, how many pairs of those points are exactly one unit distance apart? For decades, mathematicians tackled this question using various techniques, but made only incremental progress. Recently, however, artificial intelligence offered a new approach.

    A decades-old problem, solved at last

    Traditional methods for arranging points used square grids, adjusting the size to try and create more pairs of points that were a distance of 1 apart. These methods showed growth slightly faster than linear – around n multiplied by a value that barely exceeded n as the number of points increased. The general consensus became that the lowest possible growth rate was close to n^(1+o(1)), meaning just a little bit more than n, rather than a significant jump.

    How AI outperformed conjectures

    According to researchers involved, an internal model from OpenAI proposed a new family of point configurations that crosses a threshold long thought out of reach. The system produced constructions with at least n^(1+δ) unit-distance pairs, for a fixed δ greater than 0 that does not fade as n increases. That is a genuine polynomial improvement, not a rounding error.

    The solution combined an understanding of shapes with complex number theory – an unexpected combination for a problem about counting spaces. Surprisingly, it wasn’t developed by a program specifically designed for math. Instead, it came from a more general AI system being tested, which hints that this AI can think across different areas and solve problems even when there are many possibilities.

    Confirmed by experts, celebrated by the field

    Mathematicians at Princeton University independently verified the AI’s findings, according to sources close to the review. Prominent figures like Sir Tim Gowers and Arul Shankar hailed the achievement as a significant advancement in the field. Essentially, the AI broke through a longstanding barrier by discovering a new approach to a previously fixed problem.

    Implications for mathematics and beyond

    When a versatile AI model challenges long-held beliefs, it suggests a new way of working: machines propose potential solutions, and humans rigorously test them. This kind of partnership isn’t limited to geometry; fields like combinatorics, coding theory, and cryptography – where solutions depend on unusual approaches – could also benefit from this collaboration.

    Read More

    2026-06-01 00:27