This is Geoffrey Hinton’s Nobel Prize lecture, delivered on 8 December 2024 at the Aula Magna of Stockholm University and posted on the official Nobel Prize YouTube channel. Hinton shared the 2024 Nobel Prize in Physics with John Hopfield for foundational work on neural networks, and the lecture is his own account of the ideas that earned the award.
Hinton walks through Hopfield networks and then the Boltzmann machine, the stochastic learning algorithm he developed with Terrence Sejnowski and others in the 1980s. He explains how the Boltzmann machine uses a property of the Boltzmann distribution to compute the gradients needed for learning in an unexpectedly simple way, and he is widely noted for explaining the core mechanism to a general audience without writing down the mathematics.
This is a short, accessible primary source for anyone who wants to hear one of the field’s founders describe the historical roots of deep learning in his own words. It pairs naturally with the milestone entries on the 2024 Nobel Prize and on backpropagation.