Erdos Problem 1196: Can AI now solve maths that no human can?

Episode Summary

Executive Summary: The episode explores how AI appears to have solved Erdős Problem 1196, a long-standing pure mathematics problem about primitive sets. It contrasts human mathematicians’ years of work with a 23-year-old’s use of GPT-5.4 Pro to generate and verify a proof in under 80 minutes, then considers what this means for math, collaboration, and the role of human experts.

Main Topics: AI enters pure mathematics (Priority: 5/5): The episode centers on AI allegedly solving a difficult Erdős problem that had resisted human efforts for years, surprising mathematicians and raising questions about AI’s growing capabilities. What Erdős problems are (Priority: 4/5): The show explains the broader category of Erdős problems and their role in mathematics, including their popularity, scale, and the collaborative website created to track them. Primitive sets and problem 1196 (Priority: 5/5): The specific problem concerns primitive sets and how a score behaves as numbers in the set grow larger than x and x tends to infinity; the transcript notes the technical details are highly complex. Human vs AI proof verification (Priority: 5/5): The AI-generated output was initially raw and unstructured, but after review by mathematicians Jared Duker Lichtman and Kevin, it was judged to be correct, highlighting human expertise as essential for validation. Mathematical culture and collaboration (Priority: 4/5): The story frames mathematics as a field where the goal is to know whether something is true, regardless of who solves it, and suggests AI can act as a collaborative partner rather than a replacement. Historical context of unsolved problems (Priority: 3/5): Fermat’s Last Theorem and other famous conjectures are used to show how long problems can remain unresolved and why such breakthroughs matter to mathematicians.

Key Arguments: AI can now generate plausible solutions to some advanced pure mathematics problems, including ones that have resisted human experts for decades. The solution to Erdős Problem 1196 mattered because it was independently checked and recognized by a mathematician who had worked on the problem for seven years. AI output alone is not enough in mathematics; human mathematicians are still needed to interpret, verify, and understand proofs. The episode argues that in mathematics, unlike some other professions, AI’s value lies in supporting experts rather than replacing them. The achievement suggests AI may increasingly function as a brainstorming collaborator, offering intuitions and candidate arguments that humans can refine.

Data Points: AI reasoning time: about 80 minutes - Liam Price said GPT-5.4 Pro produced a candidate solution after roughly 80 minutes of thinking. Human work on problem 1196: 7 years - Jared Duker Lichtman said he had been thinking about Erdős problem 1196 for seven years. Human work on related problem 164: 4 years - Lichtman said he solved Erdős problem 164 after four years of persistence. Erdős problem collection size: 1200 problems - The transcript says the Erdős problems were collected together on a website in 2023. Date of AI breakthrough: 13 April - The episode states AI appeared to solve the problem on the 13th of April this year. Fermat’s Last Theorem resolution gap: several hundred years - Katie Steckles notes it took several hundred years before Fermat’s Last Theorem was resolved.

Pivotal Quotes: "it was pretty clear that it was correct. And the idea was very nice." — Jared Duker Lichtman: His assessment after reviewing the AI-generated proof of Erdős problem 1196. "I came up with a clever prompt and I gave it to the AI and after about 80 minutes of thinking it came out with a solution." — Liam Price: Price describing how he used ChatGPT/OpenAI’s model to generate the proof. "You're always going to need mathematicians, even if AI keeps getting better at the solutions." — Jared Duker Lichtman: His view on why human expertise remains necessary even as AI improves.

Implications: AI may become a powerful collaborator in mathematics, but human experts remain essential for interpreting and validating results. The episode suggests future breakthroughs may come from AI-assisted discovery rather than full replacement of mathematicians.