Recently, Timothy Gowers, a mathematics professor at the University of Cambridge and a Fields Medal winner, shared his experience of using the as-yet-unreleased ChatGPT 5.5 Pro in mathematical research on his blog. This AI model successfully solved an open problem in combinatorial mathematics within just one hour, sparking renewed attention in the academic community regarding AI's mathematical capabilities.
Professor Gowers pointed out that with the development of AI technology, the application of large language models in the field of mathematics has undergone significant changes. In the early days, there were doubts about their abilities, believing they could only find known answers or perform simple derivations. However, Gowers' experiments showed that ChatGPT 5.5 Pro possesses more unique capabilities, capable of identifying concise arguments that human mathematicians may have overlooked and constructing original proof ideas even without a complete theoretical framework.
The subject of this study was a problem posed by mathematician Mel Nathanson, which involved estimating the upper bound of the diameter of sum sets in additive number theory. Under Gowers' guidance, ChatGPT 5.5 Pro gradually improved from an initial exponential upper bound to a polynomial upper bound through multiple attempts, and ultimately completed the full proof within an hour. Isaac Rajagopal, an undergraduate student at MIT, who reviewed it, stated that the proof was almost certainly correct and reliable in its core ideas.
Gowers also raised an important academic ethical issue regarding how to handle these AI-generated mathematical results. Although the work done by ChatGPT has reached the standard for publication, submitting it to traditional journals seems inappropriate, so he suggested establishing a specialized repository for AI mathematical achievements, which would be reviewed by human mathematicians. This change may have a profound impact on mathematical education, graduate training, and academic evaluation systems.
Key Points:
📊 An AI model, ChatGPT 5.5 Pro, successfully cracked a Ph.D.-level mathematical problem within an hour, showcasing new mathematical abilities.
🔍 Professor Gowers pointed out that this AI not only identifies concise arguments but also proposes original ideas even without a complete theoretical framework.
📜 The academic community faces new challenges, requiring the establishment of a specialized review mechanism for AI mathematical achievements to promote the development of mathematical research.
