AI Solves 80-Year Math Problem Experts Couldn't Crack

A chatbot just accomplished what generations of brilliant mathematicians couldn't: solving an 80-year-old geometry puzzle that had experts baffled since 1946.

OpenAI announced the stunning breakthrough earlier this month when their AI model cracked the "unit distance problem," a deceptively simple question about arranging dots on paper. The challenge sounds straightforward: place nine dots so that as many pairs as possible sit exactly one inch apart.

Legendary mathematician Paul Erdős proposed the best strategy back in 1946, using a sophisticated grid approach that squeezed out slightly better results than simple patterns. For eight decades, no human could beat his record or prove he was right.

Then OpenAI researchers Mehtaab Sawhney and Mark Sellke fed the problem to their AI model trained for general reasoning. The machine churned through hundreds of pages of calculations and discovered something nobody expected: Erdős was wrong.

The AI's solution wasn't just different—it was genuinely creative. Instead of working with flat grids, the model constructed a lattice in higher dimensions with special mathematical symmetries, then cleverly mapped it back down to two dimensions.

"It feels like magic," says Sawhney. "It's kind of an amazing experience to have a machine give back something which really resembles how I work."

Independent mathematicians at Cambridge and the University of Toronto verified the proof and called it "clever" and "elegant." Timothy Gowers, a Cambridge mathematician, confirmed that no previous AI-generated proof has ever come close to meeting the standards for publication in top journals.

Why This Inspires

What makes this breakthrough special isn't just that AI beat humans—it's how the machine succeeded. The AI showed preternatural patience, exploring a difficult path that human mathematicians likely abandoned because it seemed too tedious without guarantees of success.

"AIs have an edge," says University of Toronto mathematician Jacob Tsimerman. "They can play for longer and in more treacherous waters than mathematicians without getting overwhelmed."

The achievement represents a genuine collaboration between human creativity and machine persistence. After the AI found its solution, mathematician Will Sawin already improved upon it, showing how artificial and human intelligence can build on each other's work.

The tools the AI used weren't brand new, but their application was novel. "The model did not invent something fundamentally new," says OpenAI mathematician Sébastien Bubeck. "It just executed like an amazing mathematician."

This milestone suggests we're entering an era where AI doesn't just assist with math—it can explore uncharted territories and make genuine discoveries that expand human knowledge in surprising ways.