AI Helps Crack Old Math Problems, Mathematicians Call for Careful Use

Mathematics has always been a field where human minds push the limits of logic and creativity. Recently, artificial intelligence has begun to step into this arena, showing that computers can solve problems once thought out of reach. In early May, a powerful AI from OpenAI was announced to have solved a famous geometry puzzle that had stumped researchers for eighty years. This puzzle is one of many set by the Hungarian mathematician Paul Erdos, whose questions range from simple curiosities to foundational breakthroughs. The AI’s success was not just a headline; it came with a formal paper that explained how the algorithm reached its conclusion. Some experts praised the work, saying it could be published in any top journal without hesitation.However, not everyone is convinced that the AI’s role should be uncritically celebrated. A leading mathematician from Harvard noted that the paper did not fully acknowledge previous related ideas, a concern many in the community share. The main worry is that if we rely too heavily on machines, we might lose the human insight that guides mathematical discovery. Experts suggest a balanced approach: use AI to speed up research while still ensuring that people understand the reasoning behind results. A group of sixteen mathematicians from around the world has responded by drafting a statement called the Leiden Declaration. Their goal is to set clear guidelines and spark discussion about how AI can fit responsibly into future mathematical work. The conversation is still evolving, but it is clear that both the promise and the pitfalls of AI in mathematics are now being examined with fresh eyes.