The Nobel Prize for Physics goes to the fathers of machine learning

The recognition to John Hopfield and Geoffrey E.Hinton, who first "trained" neural networks, paving the way for artificial intelligence

ROME – The 2024 Nobel Prize for Physics was awarded by the Swedish Academy to researchers John Hopfield and Geoffrey E.Hinton, who first “trained” neural networks, paving the way for “powerful automatic machine learning”, therefore to artificial intelligence.

The American John Hopfield, as we read in the note, "created an associative memory capable of archiving and reconstructing images and other types of models in the data".As for the British Geoffrey Hinton, “he invented a method that can independently find properties in data and then perform tasks such as identifying specific elements in images.”

“The work of the awardees has already brought great benefit.In physics we use artificial neural networks in a wide range of areas, such as the development of new materials with specific properties,” says Ellen Moons, president of the Nobel Committee for Physics.

The Swedish Academy continues:“When we talk about artificial intelligence, we often mean machine learning via artificial neural networks. This technology was originally inspired by the structure of the brain.In an artificial neural network, neurons in the brain are represented by nodes that have different values.These nodes influence each other through connections that can be compared to synapses and which can be strengthened or weakened.The network is trained, for example by developing stronger connections between nodes with simultaneously high values.This year's winners have conducted important work with artificial neural networks from the 1980s onwards."

The Academy continues:“John Hopfield invented a network that uses a method to save and recreate patterns.We can think of nodes as pixels.The Hopfield network uses physics that describes the characteristics of a material thanks to its atomic spin, a property that makes each atom a small magnet.The network as a whole is described in a way equivalent to the energy in the spin system found in physics, and is trained by finding values ​​for connections between nodes so that saved images have low energy.When the Hopfield network is given a distorted or incomplete image, it methodically works through the nodes and updates their values ​​so that the network's energy decreases.The network then works in stages to find the saved image that is most similar to the imperfect one it was fed with."

As for Geoffrey Hinton, “he used Hopfield's network as the basis for a new network using a different method:the Boltzmann machine.This can learn to recognize characteristic elements in a given type of data.Hinton used tools from statistical physics, the science of systems built from many similar components.The machine is trained by giving it examples that will most likely come up when the machine is started running.The Boltzmann machine can be used to classify images or create new examples of the type of pattern it was trained on.Hinton built on this work, helping to initiate the current explosive development of machine learning.”