In 2024, the Nobel Prize in Physics was awarded to American physicist John J. Hopfield and British-Canadian computer scientist Geoffrey E. Hinton, for their discoveries that enable machine learning with artificial neural networks.

 

This might sound surprising, since machine learning belongs more to the field of computer science than to physics. This contradiction is resolved by the fact that the award-winning artificial neural networks were inspired by statistical physical models. Hopfield created his memory model based on one of the simplest and most-studied statistical physics model, the so-called Ising model. The thorough understanding of Hopfield's results was made possible by methods developed in the theory of spin glasses. Hinton developed a stochastic extension of Hopfield's model, in which another cornerstone of statistical physics, the Boltzmann distribution, plays the key role.

 

Physics was not only an inspiration for creating these machine learning models, but also relies on them in applications. "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,” says Ellen Moons, Chair of the Nobel Committee for Physics.

 

Statistical phyiscs as well as machine learning plays a pivotal role BME's physics BSc, physicist-engineer BSc, and physicist MSc training, e.g., in the form of the courses Statistical Physics 1 and 2, Physics of Disordered Systems, Introduction to Machine Learning, and Artificial Intellingence in Data Science. Researchers of BME Institute of Physics apply and develop methods of machine learning: an application example is a neural-network-based method for malaria diagnostics, and a development example is the realistic simulation of a memristor-based Hopfield neural network. 

 

Press release on the webpage of the Nobel Prize: https://www.nobelprize.org/prizes/physics/2024/summary/