Hinton and Hopfield win the Nobel Prize in Physics

The Royal Swedish Academy of Sciences awarded the 2024 Nobel Prize in Physics jointly to John Hopfield and Geoffrey Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.” Neither prize concerned recent large-language-model work: Hopfield’s cited contribution was a network, developed in 1982, that could store and recreate patterns by settling into stable states, drawing on tools from statistical physics. Hinton’s cited contribution, from work through the mid-1980s, was the Boltzmann machine, a network that used those same statistical-physics tools to learn to recognise recurring elements in data — a precursor to the training methods that underlie the current generation of neural networks.

The award recognised methods roughly four decades old, honouring the theoretical groundwork rather than any specific recent model, and the Academy’s own materials framed the prize as tracing a line from 1980s statistical-mechanics techniques to the deep-learning systems that became commercially dominant after 2012.

Hinton’s award carried a specific charge given his subsequent career. He had left Google in 2023 specifically so he could speak freely about risks from advanced AI, and had spent the following eighteen months warning publicly that systems built on the foundations he helped lay could pose serious dangers, including to human oversight of increasingly capable models. He used the attention that came with the prize, including his Nobel lecture, to repeat those warnings, giving a body with no stated position on AI risk an unintended role in amplifying one of the field’s most prominent critics.