Mathematician Timothy Gowers reports a significant breakthrough with ChatGPT 5.5 Pro, which produced PhD-level mathematical research in about an hour with minimal input. The AI demonstrated the ability to identify novel solutions to complex mathematical problems, including some from the Erdős problem list, though it still relies on synthesizing existing knowledge rather than true originality.
Background
Large language models have been increasingly demonstrating capabilities in mathematical problem-solving, with recent versions showing the ability to tackle research-level mathematics problems that were previously thought to require human intuition and creativity.
- Source
- Lobsters
- Published
- May 10, 2026 at 11:01 PM
- Score
- 8.0 / 10