Skip to contentSkip to filters

We use cookies for analytics.

Human-machine mathematical collaboration with modern AI

Geordie Williamson

University of Sydney

Recording

1 h 2 min

Abstract

Geordie Williamson reflects on why computers long offered pure mathematicians little help: most interesting examples were intractable, others produced more output than anyone could interpret, and visualising results could take weeks of careful programming. Working with DeepMind, he contributed to some of the first applications of neural networks to problems in pure mathematics, where the potential was obvious but real engineering difficulties demanded close collaboration between mathematicians and engineers. Coding agents and the *Evolve family of algorithms now mark a genuine paradigm shift, making it far easier to run experiments, visualise data and connect powerful systems. Surveying his recent work, he stresses three points: the essential role of specialised software such as LP solvers, SageMath, Magma and GAP; the markedly different skill set this work requires compared with typical pure mathematics, and the challenge that poses for education at every level; and emerging problems of access to models and compute that threaten the fundamental democracy of mathematics.

Topics