John J. Hopfield of Princeton University (USA) and Geoffrey E. Hinton of the University of Toronto (Canada) have been awarded the 2024 Nobel Prize in Physics for “fundamental discoveries and inventions that enable machine learning with artificial neural networks,” the Royal Swedish Academy of Sciences announced Tuesday. Hinton was in Oviedo in 2022, where he collected the Princess of Asturias Award for Technical and Scientific Research and was also awarded the 2018 Touring Prize.
The chair of the Nobel Committee for Physics, Ellen Moons, said, “The laureates' work has already proven to be of great utility. In physics, we employ artificial neural networks in multiple areas, such as the development of new materials with specific properties.”
The two laureates have used physics principles to develop methods that underpin today's machine learning. The prize, worth 11 million Swedish kronor (approximately 1 million euros), will be shared equally between the two.
John Hopfield created an associative memory capable of storing and reconstructing images and other data patterns, while Geoffrey Hinton invented a method for autonomously identifying properties in data, such as recognizing elements in images.
The 'Hopfield network' is based on the physics of atomic spins, which turn each atom into a small magnet. This network is trained by adjusting the connections between nodes so that the stored images correspond to low-energy states. When the network receives an incomplete or distorted image, it adjusts its nodes to lower the energy and reconstruct the stored image closer to the original.
For his part, Hinton used the 'Hopfield network' as the basis for developing the 'Boltzmann machine', a network that learns to recognize characteristic patterns in data. Based on principles of statistical physics, this machine is trained on recurring examples and can classify images or generate new examples based on those patterns. This breakthrough was key to the explosive development of modern machine learning.