OpenAI AI Overturns Erdős' 1946 Conjecture After 80 Years
Updated
Updated · ScienceAlert · May 28
OpenAI AI Overturns Erdős' 1946 Conjecture After 80 Years
8 articles · Updated · ScienceAlert · May 28
OpenAI said one of its internal general-purpose AI models found a counterexample to Paul Erdős' 1946 planar unit distance conjecture, ending an 80-year search over how many point pairs can sit exactly 1 unit apart.
The proof shows infinitely many point arrangements beat the long-favored square-grid intuition by using ideas from algebraic number theory, proving Erdős' expectation was wrong.
Timothy Gowers said he would back publication in Annals of Mathematics without hesitation and called it more sophisticated than any previous AI-generated proof he had seen.
The result was produced with minimal human intervention beyond the initial prompt, and mathematician Will Sawin quickly extended the same approach to an even stronger bound.
Alongside Google DeepMind's separate resolution of 9 smaller Erdős problems, the breakthrough suggests AI is moving from mathematical assistant toward autonomous research tool, even as its ability to generate true conceptual leaps remains unclear.
Is AI's mathematical breakthrough a sign of true creativity or just superior processing power?
If an AI generates a proof too complex for humans, is it still considered knowledge?
With AI now making scientific discoveries, what is the future role for human researchers?
OpenAI’s AI Solves 80-Year-Old Erdős Geometry Problem: The Planar Unit Distance Conjecture Disproved and the Future of AI-Driven Mathematics
Overview
In May 2026, OpenAI's general-purpose reasoning AI made history by autonomously disproving the planar unit distance conjecture, a central mathematical problem posed by Paul Erdős in 1946 and unsolved for nearly 80 years. The AI achieved this by constructing an infinite family of examples, exploring solution paths that human mathematicians might have dismissed. This breakthrough, the first of its kind by an AI, was rigorously validated by the mathematical community and highlights how AI’s unique problem-solving abilities can lead to major scientific advances, opening new directions for research and collaboration between humans and machines.