John Hopfield and Geoffrey Hinton have won the Nobel Prize in Physics, 2024 for their work with artificial neural networks. They used ideas from physics to create methods that are now at the heart of today’s machine-learning technology.
How have neural networks evolved from physics research to modern machine learning?
Hopfield created a type of memory that can save and recreate pictures and other patterns. Hinton developed a way for computers to recognize details in data, like finding specific items in images.
Artificial neural networks, inspired by the brain, are made up of “nodes” that work like neurons. These nodes connect, and these connections can get stronger or weaker, helping the network learn. Hopfield and Hinton’s work in the 1980s laid the groundwork for these networks.
Hopfield’s network can remember patterns even if they are incomplete or damaged, much like how we recognise a familiar face even if the picture is blurry. He used ideas from physics, seeing each node as a small part of a larger system, with each connection working together to find the best match to a saved image.
Hinton built on Hopfield’s work to create the Boltzmann machine, which can learn to spot features in data. This system helps computers sort images and even create new patterns. His work helped start today’s rapid growth in machine learning.
Their work has had a huge impact. “We use artificial neural networks in many areas, like developing new materials,” said Ellen Moons, Chair of the Nobel Committee for Physics.
Hopfield and Hinton’s achievements remind us of the deep connection between physics and artificial intelligence, and how this relationship is shaping the future of technology.