The Nobel Prize in Physics 2024 awarded to machine learning pioneers

John Hopfield and Geoffrey Hinton have developed methods rooted in physics that have transformed artificial intelligence (AI) and, more specifically, the field of machine learning.

John Hopfield created a revolutionary associative memory system that allows machines to store and reconstruct images and patterns in data. Meanwhile, Geoffrey Hinton invented a method that autonomously identifies properties in data, enabling AI to perform tasks such as recognising specific elements in pictures.

When AI is discussed today, it often refers to machine learning using artificial neural networks, a technology initially inspired by the structure of the human brain. In artificial neural networks, the brain’s neurons are mirrored by nodes with different values, which interact through connections akin to synapses. These connections are strengthened or weakened during training, particularly between nodes with simultaneously high values. From the 1980s onward, both laureates made critical advancements in this area of study.

John Hopfield’s associative memory method allows networks to save and recreate patterns. In this network, nodes can be visualised as pixels. His model uses physics to describe the material properties based on atomic spin – a feature where each atom behaves like a tiny magnet. The network is described using energy concepts from physics, and it is trained by adjusting connections between nodes so that the stored images correspond to low-energy states. When a distorted or incomplete image is presented, the network works through the nodes to update their values and reduce the system's energy, eventually recreating a saved image similar to the one fed to it.

Geoffrey Hinton built upon Hopfield’s work, creating the Boltzmann machine, which employs a different method for recognising patterns in data. Using statistical physics – the study of systems composed of many similar components – Hinton developed a system that trains itself on data likely to arise when the machine operates. The Boltzmann machine can classify images and even generate new patterns based on its training. Hinton’s continued work in this area has fuelled the current rapid advancements in machine learning.

Ellen Moons, Chair of the Nobel Committee for Physics, highlighted the significance of their contributions: “The laureates’ work has already been of the greatest benefit. In physics, we use artificial neural networks in a vast range of areas, such as developing new materials with specific properties.”