Mathematics: the land where wild conjectures roam and theorems are painstakingly trapped by intrepid explorers armed with chalk. For centuries, this quest has been a fiercely human pursuit. But what happens when artificial intelligence takes a seat at the blackboard? Thanks to Google DeepMind’s AI, researchers are discovering new frontiers in pure maths—and we might just be living in a science-fiction story turned fact.
The Marriage of Science and Science-Fiction
- In this wild ride, we’re not just gazing at science alone. Mathematics is now joining forces with science-fiction—quite literally. The goal? To merge the best of both worlds: practical breakthroughs and the thrill of discovery that stirs our imagination.
- Here, we find not only unusual questions but also some precarious answers that spark both learning and a sense of wonder.
DeepMind Dives into Mathematical Mysteries
It’s official: artificial intelligence developed by Google DeepMind is now solving some of the trickiest problems in pure mathematics. Researchers from the University of Sydney have targeted the realms of knot theory and representation theory—fields infamous for their complexity and for leaving generations of mathematicians scratching their heads (or quietly weeping into their coffee).
This marks the first time machine learning has been used to prove new mathematical theorems, shining a light on previously unknown patterns. According to Professor Geordie Williamson, mathematician at the University of Sydney and co-author of the study, “Mathematical problems are widely considered to be some of the most intellectually challenging problems.” The practice, he says, boils down to spotting patterns and using them to formulate and prove conjectures—then, finally, theorems blossom.
Since the swinging sixties, mathematicians have relied on computers for research. Yet, as Williamson points out, machine learning only recently became a powerhouse tool for generating and analyzing complex data sets in experimental mathematics. He’s deployed DeepMind’s AI to explore conjectures in his specialty—representation theory. This area studies abstract algebraic structures by representing their elements as linear transformations of vector spaces. For the first time ever, machine learning is now crafting conjectures and suggesting possible angles to prove theorems. Mathematicians, beware—your competition is tireless and silicon-based!
Breakthroughs, Brainpower, and the Human Touch
Professor Geordie Williamson isn’t just anyone. As director of the University of Sydney’s Mathematical Research Institute and one of the most renowned mathematicians globally, he’s also the youngest living member ever elected to the Royal Society of London. He’s especially taken by representation theory—a field that handily reduces abstract algebra headaches to more accessible linear algebra ones (everyone’s favorite for a reason!).
- Crucially, representation theory saturates every corner of mathematics and has spawned countless generalizations.
- But finding new models is tougher than explaining calculus to your cat: “Working to prove or refute long-standing conjectures in my field sometimes means dealing with infinite spaces and excruciatingly complex sets of equations in multiple dimensions,” says Professor Williamson.
Williamson’s main goal? To finally crack a decades-old conjecture regarding Kazhdan-Lusztig polynomials, unsolved for 40 years. This hypothetical insight concerns deep symmetries in higher-dimensional algebra—and sounds suspiciously like the start of a math-fueled adventure novel.
His colleagues, Professor Marc Lackeby and Professor András Juhász from the University of Oxford, took things a step further. By leaning into AI, they unearthed a new connection between algebraic and geometric invariants of knots. The result? A markedly fresh theorem in mathematics. (No, not about tying your shoes—these knots reside deep in the mathematical jungle.) In knot theory, invariants are vital for telling knots apart and understanding their properties. Knot theory’s applications spill far beyond pure maths; it helps decode DNA strands, understand fluid dynamics, and even untangle the Sun’s coronal forces. Not bad for something that starts with loops of string!
When Machine Learning Meets Mathematical Intuition
But how do mathematicians typically operate? According to Professor Juhász, “Pure mathematicians work by formulating conjectures and proving them”—not exactly a Monday morning Sudoku puzzle. Most conjectures arise from mathematicians’ intuition, an elusive quality that’s particularly challenged when navigating oceans of data or confronting objects too vast for classical methods.
This is where machine learning can truly shine. These AI systems can hunt for new patterns and scenarios, simply by applying learned principles from the training data to new situations. Juhász sums it up: “When guided by mathematical intuition, machine learning offers a powerful framework to discover interesting and provable conjectures.”
The team has demonstrated that machine learning is not only useful in pure mathematics, but can actively discover models and potential relationships between objects. New and unexpected links between distinct areas of mathematics can be harnessed to lead intuition toward fresh conjectures. As Professor Lackeby puts it, their work with Oxford, Sydney, and DeepMind “shows that machine learning can genuinely be a valuable tool in mathematical research.” Williamson adds, “Intuition can take us far, but AI can help us find connections that the human mind might easily overlook.”
The authors hope their approach can serve as a model for deeper collaboration between mathematics and artificial intelligence, drawing on each discipline’s strengths to unveil surprising results. Or, as Williamson closes, the hope is that “AI can provide us with another axis of intelligence to work with, and that this new axis deepens our understanding of the mathematical world.” Math, it seems, may never be the same—and neither will we.
