In 1982, John Hopfield described a recurrent network whose stable states can store patterns as associative memories, analyzable with statistical-physics tools.
The model revived interest in dynamical networks and drew physicists into AI. Together with Boltzmann machines, it opened a language of energy, attractors, and sampling.
Though today’s dominant architectures are feed-forward nets and transformers, the Hopfield legacy returns in modern attention and memory models; Hopfield himself shared the 2024 Physics Nobel with Hinton.
