The 2023 Nobel Prize in Physics has been awarded to John J. Hopfield of Princeton University and Geoffrey E. Hinton of the University of Toronto, for their fundamental discoveries that laid the foundation for machine learning through artificial neural networks. The two scientists applied physics tools to develop innovative methods that underlie powerful machine learning technologies used today.
John Hopfield created an associative neural network
capable of storing and reconstructing images and other types of patterns within the data. This approach allowed information to be represented in a way that could be recovered even in the presence of distortions. The operation of the Hopfield network is based on physical principles that describe the characteristics of materials in relation to their atomic spinor, a phenomenon that gives each atom the property of a small magnet. The network is trained to optimize connections between nodes, ensuring that stored images have minimal energy. When the network is fed a distorted or incomplete image, it processes the nodes and updates their values until it finds the saved image that is closest to the imperfect one.
Geoffrey Hinton took inspiration from Hopfield’s network to develop a new network architecture known as a Boltzmann machine. This model can learn to recognize characteristic elements within a specific data type. Hinton applied tools of statistical physics, the branch of science that studies systems composed of multiple similar components. The training of the Boltzmann machine occurs by feeding highly probable examples, allowing the network to classify images or generate new examples of patterns on which it has been trained.
Hopfield and Hinton’s work has already proven to have a significant impact, not only in the field of artificial intelligence, but also in applied physics. Ellen Moons, president of the Nobel Physics Commission, commented: “The work of the winners has already brought enormous benefits. In physics, we use artificial neural networks in a wide range of areas, such as in the development of new materials with specific properties”.
The discoveries of these two scientists laid the foundation for the current explosion of machine learning development, showing how the integration of different scientific disciplines can lead to innovations that transform our approach to technology and understanding the world.
Beyond the academic recognition, the two winners will also receive the financial prize of 11 million Swedish crowns, just under one million euros, to be shared.
image credits afp
#John #Hopfield #Geoffrey #Hinton #Nobel #Prize #Physics #discoveries #machine #learning