AI models achieve breakthroughs in complex mathematical research

Artificial intelligence is making significant strides in complex mathematics, evolving from a simple problem-solving tool into a potential research partner for scientists. OpenAI reported that an internal version of its Astra model achieved notable results in ten problems involving group theory, geometry, cryptography, and quantum theory—areas where progress has been stalled for decades. While these results include both complete solutions and significant advancements, they were refined by humans and formalized using Lean software to ensure mathematical validity.

Similarly, Anthropic’s Claude model has shown progress regarding the Riemann Hypothesis. The model improved the known lower bound for the percentage of zeros of the Riemann zeta-function on the critical line, increasing it from 41.6 percent to 67 percent. Despite these breakthroughs, some members of the mathematical community have raised concerns regarding the clarity of the results and how they distinctly differ from previous research.