INI congratulates John Hopfield and Geoff Hinton for their award of the 2024 Nobel Prize in Physics.
The Institute of Neuroinformatics owes a great thanks to John Hopfield. His beautiful 1982 and 1984 PNAS papers on collective computation in recurrently-connected content-addressable associative memory ignited interest in the physical foundations of neural computation. His course (later CNS182) on the physics of computation taught with Dick Feynmann and Carver Mead and his subsequent work to form the CNS program at Caltech brought together INI founding members Rodney Douglas, Misha Mahowald, Jorg Kramer, Giacomo Indiveri, Shih-Chii Liu, and Tobi Delbruck, who shared the vision of INI as a place to connect biology, physics, and engineering to understand and implement neural computation.
Hopfield got interested in neural computation from the viewpoint of a biophysicist who had his 1960’s roots in solid state physics and mechanisms of electron transport in molecules. In the early 1980’s, machine learning was mired in a feeling that it could only do clustering, classifiers that linearly separated points in hyperspace, and a range of other ad hoc developments like adaptive resonance theory. Hopfield’s 1982 and 1984 PNAS papers ignited interest in neural computation because they clearly and concisely showed that it was possible to formulate memory as energy minimization and provided a simple recipe to store and retrieve memories and even proposed a concrete and simple electronic circuit implementation. It did not matter that the Hopfield network had no immediate application because it helped to inspire a generation of researchers.