AI Breakthrough in Mathematics: The Erdős Conjecture

OpenAI researchers recently utilized a general-purpose large language model to successfully disprove the 80-year-old unit distance problem, a famous conjecture proposed by mathematician Paul Erdős in 1946. The AI achieved this by constructing a complex counterexample using tools from algebra and number theory, marking a significant milestone in AI-assisted scientific discovery. This breakthrough demonstrates the potential for AI to apply cross-disciplinary mathematical tools to solve long-standing geometric challenges. Despite the achievement, experts emphasize that the AI relied on computational perseverance rather than creative insight, leading to a broader debate regarding the future of mathematical research. On June 2, 2026, a group of mathematicians issued a declaration calling for strict guardrails to ensure that AI-generated proofs remain verifiable, ethical, and transparent. The academic community remains cautious, balancing optimism for AI as an indispensable tool with concerns over the lack of attribution, potential for unreliable reasoning, and the risk of eroding democratic access to mathematical knowledge. The core issue lies in the opacity of AI models, which often fail to show the work or logic behind their conclusions, making it difficult for human experts to verify complex proofs. Furthermore, the lack of transparency regarding failure rates and the proprietary nature of these powerful models pose significant challenges to the standard practices of mathematical verification and credit attribution. For researchers and practitioners, this event highlights the necessity of developing rigorous verification protocols for AI-generated results. Users should approach AI-assisted proofs with skepticism, prioritizing human-led verification and advocating for open-access, transparent AI tools to maintain the integrity and collaborative spirit of the mathematical community.