AI Solves Erdős Problems at an Unprecedented Rate—But Can It Outpace Human Genius?

A software engineer's experiment with ChatGPT 5.2 has cracked open one of math’s most stubborn unsolved problems—just in time for the AI revolution to outpace human mathematicians.

Neel Somani noted the significance of establishing benchmarks for AI in mathematical problem-solving:

"I was curious to establish a baseline for when LLMs are effectively able to solve open math problems compared to where they struggle," said Neel Somani

Tudor Achim highlighted a broader shift in academic adoption:

"I care more about the fact that math and computer science professors are using AI tools."

Since Christmas 2023, 15 Erdős problems have been marked as solved, with 11 crediting AI models. Terence Tao's GitHub analysis reveals 8 problems where AI made autonomous progress and 6 where AI built on prior research. GPT 5.2 shows anecdotally observed improvements over previous iterations in mathematical reasoning.

Tools like Harmonic's Aristotle and OpenAI's deep research initiatives are formalizing proofs and automating literature reviews. Microsoft Research's Proof assistant Lean (2013) has accelerated formalization of mathematical reasoning, enabling more rigorous validation of AI-generated solutions.