A Knotty Problem
One of the most compelling examples of this new partnership comes from the abstract field of knot theory. Mathematicians in this area study the properties of knots, much like the ones you’d find in a shoelace, but in a purely mathematical sense. A key
challenge is distinguishing one knot from another and understanding their fundamental properties. For decades, mathematicians suspected a hidden relationship between two different ways of describing knots: their 'algebraic' properties and their 'geometric' properties. While the intuition was there, proving a concrete connection was a formidable task that stumped researchers for years.
The AI as an Intuition Engine
Researchers from DeepMind, a subsidiary of Google's parent company Alphabet, collaborated with mathematicians to tackle this challenge. They trained a machine learning model to analyze a vast database of knots and their known properties, or 'invariants'. The AI's job wasn't to perform brute-force calculations, but to hunt for subtle patterns and relationships that were not obvious to the human eye. By feeding the AI data on thousands of knots, the system began to predict one type of property based on the other with surprising accuracy. This suggested a deep, underlying connection was indeed present.
A Human and Machine Partnership
Crucially, the AI did not 'solve' the problem on its own. Instead, it acted as a guide for human intuition. The patterns it identified pointed the mathematicians in the right direction, highlighting which specific properties were most strongly related. Armed with this AI-generated insight, the mathematicians were able to formulate a new conjecture—a proposed theorem—linking the two types of knot invariants. One of the mathematicians involved, Geordie Williamson of the University of Sydney, noted that the AI models supported the intuitive and creative aspects of mathematics in a way he hadn't experienced from computers before. This collaboration ultimately led to a brand new, human-proven theorem in the field.
Why This Changes Mathematical Research
This breakthrough represents a significant shift in how scientific discovery can happen. The AI wasn't just checking work; it was generating hypotheses and providing a new kind of intuition. Problems that are too complex or have too many variables for a human to hold in their mind can be sifted by an AI, which can spot non-obvious correlations. This approach is already being applied elsewhere. In other recent events, AI has been used to make progress on a 40-year-old problem in a field called representation theory and has even disproven a long-standing conjecture from the famous mathematician Paul Erdős. In these cases, the AI often finds a novel approach that human experts had overlooked for decades.
