OpenAI's AI Achieves Breakthrough in Erdős Unit Distance Problem

OpenAI announced a breakthrough in mathematics: one of its general-purpose AI models solved a famous discrete geometry problem nearly 80 years after it was posed by Hungarian mathematician Paul Erdős. According to the company, the result has already undergone independent expert review by the mathematical community.

The problem in question is the so-called unit distance problem on the plane. It explores the maximum number of point pairs that can be arranged so that the distance between them is exactly one unit. Since the mid-20th century, it was believed that the most efficient configurations would be close to a classical square grid, but a definitive proof had never been achieved.

OpenAI says the model approached the problem in an unexpected way, using ideas from algebraic number theory instead of the usual geometric methods. As a result, it proposed a new family of point configurations that yields more valid combinations than previous estimates suggested.

Another notable aspect is that the model was not specifically designed for solving mathematical problems. The company emphasizes that a general reasoning system was used, rather than a specialized mathematical tool.

Several independent mathematicians reviewed the proof and confirmed the validity of the core idea. Researchers note that this could be an important example of how AI is not only capable of accelerating the analysis of existing solutions but also of proposing new directions for scientific inquiry. However, the problem itself is not considered fully solved—further questions about the exact boundaries of the result remain open.

Interestingly, similar claims have sparked controversy in the past: there have been reported cases where AI found known solutions rather than creating new ones. This time, OpenAI specifically emphasized that it underwent external verification and that the research materials have been published.