The physical origins of artificial intelligence

Speaker: Surya Ganguli (Stanford University) In Person and zoom

Zoom Link

Colloquium Webpage

We will review the science underlying the recent award of the Nobel Prize to John Hopfield and Geoffrey Hinton for developing the foundations of modern artificial intelligence using tools from physics.  This award has generated some misplaced controversy, with (some younger) computer scientists claiming AI has nothing to do with physics, and (some older) physicists claiming physics has nothing to do with AI.  On the contrary, I will explain how this prize was an inspired choice awarding decades of ground breaking interdisciplinary work originating in statistical physics, and leading through a direct causal chain to modern deep learning, through seminal models like the Hopfield model and the Boltzmann machine. I will also explain how the Hopfield model is connected to very recent developments, including diffusion models and transformers. Throughout, I will also sprinkle a few jokes I personally exchanged with Geoffrey Hinton last week.