AI Cracks Erdős's Planar Unit Distance Problem for the First Time: A Landmark Breakthrough in Mathematical History

A Historic Moment: AI Solves a Truly Famous Mathematical Problem

AI achievements in mathematics are nothing new, but this time the significance is fundamentally different. A research team has announced that AI achieved a major breakthrough on the Planar Unit Distance Problem — a classic open problem first posed by legendary mathematician Paul Erdős, widely regarded as the most famous unsolved problem in combinatorial geometry.

Participants described the achievement this way: "This is the first time AI has clearly and definitively solved a mathematical problem that is not just unsolved, but widely known to be unsolved. This is the first mathematical breakthrough attributable to AI."

Erdős and Combinatorial Geometry: A Century-Old Challenge Left by a Legendary Mathematician

Paul Erdős (1913–1996) was one of the most prolific mathematicians of the 20th century, publishing over 1,500 papers with collaborators worldwide, giving rise to the famous "Erdős number" concept — a measure of collaborative distance between a mathematician and Erdős. He was particularly renowned for posing problems that seem deceptively simple yet prove extraordinarily difficult to solve, offering cash prizes to incentivize the mathematical community to tackle them.

Combinatorial geometry is a branch of mathematics that studies geometric properties of finite point sets, sitting at the intersection of discrete mathematics and classical geometry. The series of problems Erdős posed in this field form the core research agenda of the discipline, and the Planar Unit Distance Problem is considered one of the most difficult and far-reaching among them — ranked alongside the Erdős–Szemerédi conjecture and the Erdős distinct distances problem as the three great classic problems of combinatorial geometry.

The Erdős Planar Unit Distance Problem: Simple Enough for a Child to Understand, Hard Enough to Stump Mathematicians

The Planar Unit Distance Problem has an extremely concise statement — it concerns points in the plane and the distance relationships between them: Given n points in the plane, what is the maximum number of pairs of points that are exactly distance 1 apart (i.e., at "unit distance")? This is a completely elementary geometric question — anyone can understand what the problem is asking.

In mathematical notation, this maximum is denoted u(n). Erdős himself conjectured that the upper bound of u(n) is approximately n^(1+c/log log n), while the best known lower bound construction (given by Hungarian mathematicians through special lattice point constructions) is approximately n^(1+c/√log n). The enormous gap between the upper and lower bounds, persisting for decades, is precisely the core difficulty of this problem.

However, solving this problem requires mathematical tools of an entirely different caliber. Previously, the mathematical community had a construction considered to be "essentially optimal," and the consensus was that this construction was already near the limit and could not be significantly improved.

The AI model shattered this consensus — it proved that the existing construction can in fact be substantially improved, and by a considerable margin.

Why Could AI Achieve What Humans Could Not?

The most thought-provoking aspect of this breakthrough is that the core idea of the solution is not entirely beyond human comprehension, but humans cannot actually execute this approach.

Researchers explained the key reasons behind this:

• The solution involves very deep tools from algebraic number theory
• The proof requires making a vast number of delicate decisions and choices
• Each step's choices are extremely subtle ("too delicate")
• The number of possible path combinations is enormous, far exceeding human exhaustive capacity

Algebraic Number Theory: The Deep Mathematical Structure Behind the Proof

Algebraic number theory is a field at the intersection of number theory and abstract algebra, studying the applications of algebraic integers, ideals, field extensions, and other structures to problems in number theory. In the proof of the unit distance problem, key tools involve lattice constructions over algebraic number fields, special algebraic structures such as Gaussian integers and Eisenstein integers, and the norm distribution properties of elements within these structures.

Specifically, when seeking large numbers of unit-distance pairs in the complex plane, the problem essentially transforms into: within the ring of integers of a certain algebraic number field, how many pairs of elements have a difference whose norm equals 1? This transformation maps the geometric problem onto algebraic structures, but simultaneously introduces an exponentially growing parameter selection space — this is precisely where humans cannot exhaustively search, but AI can systematically explore.

AI's advantage lies precisely in its ability to systematically explore all these possibilities. Where humans cannot make progress due to an overwhelmingly large decision space, AI searched the entire proof path space and ultimately found the route to the solution.

Technical Pathways for AI Mathematical Reasoning

The methods employed by AI in this breakthrough are closely related to recent technical advances in large language models for mathematical reasoning. Current state-of-the-art mathematical AI systems typically combine multiple technical approaches: formal proof verification (using proof assistants like Lean and Coq), reinforcement learning-driven proof search (similar to AlphaGo's tree search strategy in board games), and mathematical pattern recognition capabilities derived from large-scale pretraining.

At the path search level, AI's advantage lies in its ability to prune and explore high-dimensional decision trees without abandoning seemingly tedious but actually viable paths due to cognitive overload, as humans tend to do. This belongs to the same technical paradigm as DeepMind's AlphaTensor, which discovered new matrix multiplication algorithms in 2022 — this is not brute-force computation, but rather a demonstration of superhuman exploration capability in domains requiring deep mathematical intuition and fine-grained judgment.

Mathematicians' Genuine Reactions: Shock and Sleepless Nights

The reactions of mathematicians involved in verifying this result were remarkably visceral:

"When I saw the initial version of the model's output, I simply couldn't believe it. I spent a long time reading it over and over, trying to figure out what had actually happened. I thought: you can't be serious. This sounds too good to be true."

"I simply couldn't believe it. I couldn't sleep for the first two nights. I was completely stunned."

Another researcher admitted: "I always expected AI to do something like this, but this result drastically shortened my timeline." — AI's capability progress far exceeded his expectations.

The team initially applied the model to just a few Erdős problems as a test. "To our surprise, it came back with a solution to one of the most important Erdős problems."

Beyond Mathematics: A New Paradigm for AI-Driven Scientific Breakthroughs

The significance of this result extends far beyond solving a single mathematical problem. Researchers explicitly pointed to broader implications:

"This tells us that AI can achieve breakthroughs in science — not just mathematics, but also engineering, physics, biology, and medicine."

The Historical Evolution of Scientific Breakthrough Paradigms

Throughout the history of science, revolutions in research tools have often preceded explosions of major discoveries. The invention of the telescope ushered in the astronomical revolution, the advent of computers gave birth to computational physics and computational biology, and the internet reshaped scientific collaboration. AI, as the next generation of scientific research tools, is unique in that it not only accelerates computation but can also participate in creative work at the level of abstract reasoning.

Previous landmark AI achievements in science include: AlphaFold solving the protein folding problem (2020), AlphaTensor discovering new matrix multiplication algorithms (2022), and AI-assisted discovery of new antibiotics (2023). What distinguishes this mathematical breakthrough is that it solved a "named problem" with clear historical documentation that has been the focus of sustained attention from the mathematical community.