Over the weekend, software engineer Neel Somani explored the math capabilities of OpenAI’s latest model, discovering its ability to solve complex problems, including those from famed mathematician Paul Erdős. After testing ChatGPT, Somani confirmed a full solution using the tool Harmonic, which evaluates mathematical proofs. He noted that the latest iteration, GPT 5.2, demonstrates enhanced mathematical reasoning skills. AI tools like Harmonic are increasingly integral in tackling Erdős’s over a thousand conjectures, with 15 problems recently classified as “solved,” 11 of which credited AI assistance. Renowned mathematician Terence Tao acknowledged AI’s potential in systematically addressing obscure mathematics problems, citing AI’s efficiency. The shift towards formalization in mathematical proofs, facilitated by tools like Lean, further underscores AI’s role in mathematics. As more top mathematicians adopt these technologies, the integration of AI in solving complex mathematics becomes increasingly credible, pointing to a transformative future in mathematical research.
Source link
