23-Year-Old Student Solves 60-Year-Old Math Problem by Asking AI
A 23-year-old student has solved an unsolved mathematical problem that stumped top mathematicians for 60 years — simply by asking ChatGPT. A landmark achievement showing AI's potential to open new frontiers in mathematics.
23-Year-Old Student Solves a 60-Year-Old Unsolved Math Problem with a Single AI Query
For over 60 years, some of the world’s top mathematicians had been unable to crack a particular mathematical problem. Then a 23-year-old student solved it — with a single question to an AI. The feat was accomplished by Ryan Price.
Breaking a 60-Year Barrier with a “Simple” Prompt
Price posed a question to ChatGPT about one of the Erdős problems — a collection of unsolved mathematical problems posed by the legendary mathematician Paul Erdős, many of which have remained unresolved for decades.
Price used OpenAI’s latest model, “GPT-5.4 Pro.” He presented the problem to the AI and asked for a solution. The result: the AI generated a proof using a method that humans had never discovered before.
The solution was posted on the mathematical community website erdosproblems.com, drawing significant attention from mathematicians worldwide.
AI Opens a New Era in Mathematical Research
Traditionally, mathematical proofs have been considered products of human intuition and creativity. This event, however, suggests that AI can independently arrive at mathematical truths through its own approaches.
What is particularly striking is that the proof technique the AI employed had “not been developed by humans.” This indicates that AI can not just reproduce human knowledge but can also offer genuinely novel mathematical insights.
The Impact of Amateur + AI
Price is not a professional mathematician. Yet, by skillfully leveraging the latest AI tools, he found a path to a solution that had eluded experts for years.
This case strongly points to AI’s potential to democratize research. The era is arriving when even those without extensive specialized expertise can contribute to cutting-edge research, provided they know how to use the right AI tools effectively.
Community Reaction and Future Challenges
The mathematical community is now in the process of verifying this solution. AI-generated proofs often have logical structures that differ from traditional proofs, and new criteria for evaluating their validity may be needed.
The limits of AI’s problem-solving capabilities in mathematics, as well as the ideal form of collaboration between human mathematicians and AI, will likely be subjects of deeper discussion going forward. Price’s landmark achievement may fundamentally reshape how we think about the relationship between AI and human intellectual endeavor.
Frequently Asked Questions
- To what extent can AI's ability to solve math problems be expected to improve in the future?
- Current AI models show excellent performance on certain types of mathematical problems, but they have not yet reached a level comparable to human mathematicians in general mathematical reasoning. However, as computational power improves and training data expands, they may become capable of tackling increasingly complex problems. This case suggests that AI can not only serve an assistive role in mathematical research but also provide original insights — its future evolution will be closely watched.
- Are AI-generated proofs as trustworthy as those produced by human mathematicians?
- AI-generated proofs are considered mathematically correct as long as their logical consistency is maintained. However, like human proofs, they require verification by other researchers. When AI uses methods that differ from conventional approaches, new validation processes may be needed to assess their validity. In this case as well, ongoing verification by the mathematical community is underway.
Comments