Deepmind AI detects new sports links

Photo: DeepMind

In Deepmind’s latest research, an AI system in collaboration with humans is revealing links previously unknown in mathematics.

An artificial intelligence system from British AI company Deepmind has for the first time revealed mathematical connections that mathematicians had previously hidden. For the math project, Deepmind worked with three mathematicians to solve two math problems – one in knot theory and one in symmetry theory.

As inspiration for the project, Deepmind cites Indian mathematician Srinivasa Ramanujan, who is credited with an extraordinary ability to see patterns in numbers. He himself described his findings as self-evident and often saw patterns in vivid dreams. Similar to how a mathematical genius should Artificial intelligence Detecting unknown patterns in the data.

Such data exists thanks to the use of computers in experimental mathematics, which has already given rise to many mathematical conjectures. These assumptions are statements that experts assume are true. Once established, mathematics searches for a proof or refutation.

Deepmind’s AI studies knots and symmetries

With the help of artificial intelligence at Deepmind, mathematician Jordi Williams of the University of Sydney has found a new approach to estimating the 40-year-old combinatorial constant, which states that some vector and polynomial graphs must be related to each other. The use of artificial intelligence has shown that such a relationship does exist and has identified the structures with which this relationship is connected.

With mathematicians Mark Lackenby and Andras Johasz of the University of Oxford, DeepMind investigated knots, a fundamental subject of study in topology. Deepmind’s AI system searched for patterns in different perspectives of algebra, geometry and quantum theory on nodes and showed that an algebraic quantum “signature” is directly related to node geometry. With the help of artificial intelligence, Professor Lackenby also discovered the new variable “normal slope” and established the relationship between the two variables.

Participating mathematicians are positive about AI in mathematics

If Deepmind has its way, machine detection of patterns and generation of guesses should become the norm in mathematics in the future. The participating mathematicians are positive about collaborating with AI.

“You might think a mathematician’s work is dry and sensible. The reality is quite different. Mathematicians live in a world rich in imagination, reasoning, and intuition,” Williamson said.

The collaboration was a fascinating interdisciplinary journey and showed that AI can be used to control intuition and prove new theories. On the Numberphile podcast, Alex Davies of Deepmind and mathematician Marcus du Sautoy talk about the latest successful AI research.

Lackenby is also convinced: “I was amazed at how useful machine learning tools were as guidelines for intuition,” says the mathematician. He didn’t expect some of his prejudices to be turned upside down.

Results and procedures describe all participants in a scientific paper in Nature Magazine. There it is said that the case study showed that an essential link in a well-studied and exciting field of mathematics can go unnoticed and be discovered through human-machine collaboration.

The new results will be published in specialized sports journals. Deepmind also publishes associated AI models github.

Read more about Deepmind: