Liam Price uses ChatGPT Pro to solve longstanding Erdős mathematical problem

4 articles · Updated · Scientific American · Apr 24

At age 23, Price prompted GPT-5.4 Pro to solve a 60-year-old Erdős problem regarding primitive sets, posting the solution online last week.
Experts, including Terence Tao and Jared Lichtman, confirmed the AI’s novel approach, which applied a known formula in an unexpected way and may have broader mathematical applications.
The solution’s originality distinguishes it from recent AI-generated proofs, fueling optimism about AI’s potential in mathematics, though experts caution the long-term significance remains to be seen.

How did a math amateur and an AI solve a problem that stumped geniuses for 60 years?

What was the AI's secret method that eluded the world's best mathematicians for decades?

What other 'unsolvable' scientific problems could next fall to AI's novel way of thinking?

What was the single prompt that unlocked a 60-year-old mathematical mystery?

Is this the end of human mathematical intuition or the start of a powerful new partnership?

If AI's raw output was 'poor,' who deserves credit for the discovery: human or machine?