OpenAI model disproves Erdős’s unit-distance conjecture (discrete geometry)
OpenAI model disproves Erdős’s unit-distance conjecture (discrete geometry)
OpenAI reports that an internal general-purpose reasoning model disproved a central conjecture about the planar unit distance problem (posed by Paul Erdős, 1946): how many unit-distance pairs can appear among n points in the plane.
The result is notable both mathematically (a polynomial improvement construction, via ideas from algebraic number theory) and historically: OpenAI describes it as the first time a prominent open problem central to an active subfield has been solved autonomously by AI, with the proof checked by external mathematicians.
Key quote (Tim Gowers)
“There is no doubt that the solution to the unit-distance problem is a milestone in AI mathematics: if a human had written the paper and submitted it to the Annals of Mathematics and I had been asked for a quick opinion, I would have recommended acceptance without any hesitation. No previous AI-generated proof has come close to that.”
Links
- OpenAI overview/announcement: https://openai.com/index/model-disproves-discrete-geometry-conjecture/
Why this matters
A genuinely big milestone: AI proving a significant, long-standing open problem in mainstream mathematics (not just assisting), and doing so with techniques that connect distant areas (algebraic number theory → discrete geometry).