OpenAI and DeepMind Just Broke an 80-Year-Old Math Problem — And Solved 9 More That Stumped Experts for Decades

In a stunning week for mathematics, OpenAI and Google DeepMind have demonstrated AI's ability to tackle iconic open problems. OpenAI's model disproved the planar unit distance conjecture posed by Paul Erdős in 1946, a problem that mathematicians had assumed was best solved by square grids. DeepMind followed by solving nine problems, including two that had remained unsolved for over 50 years. Carnegie Mellon professor Jeremy Avigad called it a 'considerable mathematical achievement' and predicted this would not be the last such case. Brown University professor Javier Gómez Serrano, who collaborated with Google on Navier-Stokes equations, described the moment as 'a real inflection point.'

The breakthroughs have triggered mixed reactions among mathematicians—fear, indifference, and even a manifesto. The Leiden Declaration on Artificial Intelligence and Mathematics now recommends disclosing AI use in scientific articles and confirming human authorship. While some worry about 'the end of mathematics,' experts like Petra Schwer of Heidelberg University see AI as a tool similar to calculators or computers. DeepMind founder Demis Hassabis cautioned that today's systems are 'extremely far' from true invention like Ramanujan. Nonetheless, the attention-grabbing headlines are fueling investment in AI-driven math startups, raising ethical questions about credit and collaboration in a field built on rigorous norms.