AI uncovers solutions to Erdős problems, moving closer to transforming math (www.scientificamerican.com)

A recent breakthrough by mathematicians using AI tools has led to the resolution of several long-standing problems left by the renowned mathematician Paul Erdős. Mehtaab Sawhney from Columbia University, leveraging ChatGPT, discovered that the AI could connect him to existing references and solutions for Erdős problems. Collaborating with fellow mathematician Mark Sellke from OpenAI, the duo prompted the AI to uncover solutions for nine problems and partial solutions for eleven others. This marked the beginning of a wave of activity where AI has reportedly helped solve around 100 Erdős problems, showcasing its evolving role as a powerful mathematical search engine capable of synthesizing literature and assisting in problem-solving. The implications for the AI/ML community are significant, indicating a shifting landscape in mathematical research. While AI cannot yet replace human mathematicians, it is proving to be a valuable research assistant, sparking excitement and encouraging discussions about its potential. Current challenges include verifying the correctness of AI-generated solutions, as many claims are found to be inaccurate. With initiatives like "First Proof," which tests AI's ability to resolve unsolved mathematical proofs, the mathematicians are exploring the integration of AI into their work. Experts predict that within a few years, AI’s contributions could influence peer-reviewed publications, potentially transforming the nature of mathematical inquiry and collaboration.