OpenAI revealed that its newest AI model has effectively refuted the unit distance conjecture, a question that has lingered unsolved for roughly 80 years. The conjecture, first posed by mathematician Paul Erdős, asks whether there exists a set of points in the plane such that the number of pairs at exactly one unit apart can be maximized beyond a certain bound. While the problem has inspired countless papers, no concrete counterexample had emerged—until now.

The breakthrough stems from a clever manipulation of a regular grid. By shrinking the spacing between points, the AI discovered that a unit‑distance circle can intersect many more grid points than the classic arrangement allows. In one illustrative diagram, the model set the grid spacing to 1/√65, producing a circle that would intersect sixteen points if the grid were extended far enough. This configuration eclipses the earlier best‑known example, which yielded only twelve intersecting points.

Underlying the discovery is the Pythagorean theorem, which links horizontal and vertical offsets to the straight‑line distance between two points. The AI selected a specific squared distance, denoted c², that admits a large number of integer solutions to a² + b² = c². For instance, choosing c² = 25 yields the familiar (3,4,5) triple, but the model identified a more fertile value that generates a dense set of integer pairs. By scaling the grid so each point lies 1/c units from its neighbors, the system created a lattice where many neighboring points sit exactly one unit apart.

OpenAI’s write‑up includes a series of animated diagrams that show the “distance‑one neighbors” for central points in a 13×13 grid. When the grid spacing is set to one, each point touches only four neighbors—up, down, left, and right. Reducing the spacing to one‑fifth expands that neighborhood to twelve points, and the AI’s optimal spacing pushes it even further. The visualizations help demystify a result that, on paper, could appear abstract and inaccessible.

Experts who reviewed the findings say the AI’s approach mirrors traditional mathematical reasoning while leveraging brute‑force computation at a scale humans cannot easily replicate. The model systematically explored a vast space of possible grid spacings and distance values, testing each configuration against the unit‑distance condition. When a promising candidate emerged, the system refined it, ultimately arriving at the 1/√65 spacing that yields the unprecedented sixteen‑point intersection.

The result has sparked excitement across both the AI and mathematics communities. It showcases how machine‑learning systems can generate novel conjectures, construct counterexamples, and even suggest new avenues of proof. While the unit distance conjecture is now considered disproven, the broader implication is that AI may soon assist in resolving other long‑standing open problems, from prime number distribution to topology.

OpenAI plans to release the code and detailed methodology behind the discovery, inviting peer verification and further exploration. If the community validates the counterexample, textbooks will need to be updated, and a chapter of mathematical history will close with an AI‑driven proof. The episode underscores a shift: artificial intelligence is moving from pattern‑recognition tasks to the front lines of pure theoretical research.

This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.