Artificial intelligence is making remarkable progress on some of the most famous open problems posed by the legendary Hungarian mathematician Paul Erdős, surprising researchers who once believed these challenges would remain beyond AI's reach for years. According to Quanta Magazine, several recent AI breakthroughs—including systems from OpenAI and Google DeepMind—have successfully solved or advanced multiple Erdős problems in combinatorics and graph theory. These achievements have prompted mathematicians to rethink AI's role, not just as a computational assistant but as a genuine collaborator in mathematical discovery.
The article explains that Erdős' problems are especially well suited to AI because they often involve searching enormous combinatorial spaces for elegant patterns, constructions, or counterexamples. Unlike many areas of mathematics that rely on deep conceptual intuition, these problems can sometimes be approached through systematic exploration combined with logical reasoning. Modern AI systems excel at navigating vast search spaces, testing hypotheses, and combining symbolic reasoning with machine learning, enabling them to uncover solutions that human mathematicians had not considered.
Researchers emphasize that AI is not replacing mathematicians but changing how mathematical research is conducted. Human experts remain essential for formulating meaningful questions, interpreting results, verifying proofs, and identifying broader mathematical significance. AI-generated proofs are increasingly being checked using formal verification systems, while mathematicians work to understand the new techniques AI introduces. Some AI-derived proofs have employed approaches that differ significantly from traditional human reasoning, creating opportunities for entirely new mathematical insights.
The broader significance extends well beyond combinatorics. The article suggests that Erdős' problems may represent an early glimpse of how AI could transform mathematics as a whole. As reasoning models become more capable and formal proof systems continue to mature, AI is expected to become an increasingly powerful research partner across many branches of mathematics. At the same time, these breakthroughs have sparked discussions about transparency, verification, attribution, and the future role of human creativity, as the mathematical community works to ensure AI strengthens—rather than replaces—the collaborative and rigorous nature of mathematical discovery.
