John J. Hopfield

Overview

John J. Hopfield is an American physicist, born on 15 July 1933 in Chicago, Illinois^[1][2]. He is best known for a model of neural computation he invented in 1982: a network that uses a method for saving and recreating patterns^[1]. In 2024 he was awarded one half of the Nobel Prize in Physics for foundational discoveries and inventions that enable machine learning with artificial neural networks^[1]. At the time of the prize he was a professor at Princeton University in New Jersey^[1][2].

Education and early career

Hopfield received a BA in physics from Swarthmore College in 1954 and a PhD in physics from Cornell University in Ithaca, New York, in 1958^[3][2]. He spent the early part of his career in physics departments. He was on the physics faculty at the University of California, Berkeley, from 1961 to 1964, and then on the physics faculty at Princeton University from 1964 to 1980^[3]. His early recognition came from condensed matter physics: in 1969 he received the Oliver E. Buckley Prize for condensed matter physics, shared with David Gilbert Thomas^[3], and he was elected to the National Academy of Sciences in 1973^[3].

The Hopfield network

In 1982 Hopfield invented a network that uses a method for saving and recreating patterns^[1]. The idea came from outside computer science. The Hopfield network borrows the physics that describes a material's properties arising from atomic spin^[2]. Hopfield drew the idea from physics models of how many small parts in a system affect the system as a whole^[1].

The network is described in the language of that physics. Its state is written in terms equivalent to the energy of a spin system, and it is trained by setting the connection values so that the stored images sit at low energy^[2]. Recalling a pattern is then a matter of letting the system slide downhill. Given a distorted or incomplete image, the network updates node values stepwise so that the network's energy falls, until it settles on the stored pattern closest to the input^[2]. In this way the same mathematics that describes how spins in a magnet settle into a stable arrangement is used to clean up and recover a memory.

The invention became important in applications such as image analysis^[1]. It also became a starting point for other researchers. Geoffrey Hinton used the Hopfield network as the foundation for a new network built on a different method^[2].

From physics to biology

In 1980 Hopfield moved to the California Institute of Technology, where he was a professor of chemistry and biology from 1980 to 1997^[3]. He returned to Princeton University in 1997 as Howard A. Prior Professor of Molecular Biology^[3], and he was still a professor at Princeton in New Jersey when the 2024 prize was announced^[1][2].

Recognition and influence

Hopfield received the IEEE Neural Networks Pioneer Award in 1997^[3]. The 2024 Nobel Prize in Physics recognised foundational discoveries and inventions that enable machine learning with artificial neural networks; Hopfield received one half of the prize^[1]. His contribution sits at the meeting point of two fields: a physicist's way of thinking about many interacting small parts, applied to the problem of storing and recovering patterns in a network of simple units. That framing — a network with an energy that falls as it computes — gave later work in neural networks both a vocabulary and a working model to build on^[2].