PERSON
John Hopfield
The physicist who built memory from the mathematics of magnets—his 1982 network modeling recall as descent through an energy landscape toward stable attractors—and whose founding contribution to artificial intelligence is inseparable from his physicist’s dread of a powerful technology no one fully understands.
In 1982 John Hopfield published a five-page paper proposing that memory could be a
landscape: a network of simple, neuron-like units could store recollections as valleys in an energy terrain and recover them by rolling downhill into the nearest one. Borrowed from the physics of spin glasses—disordered magnets whose conflicting couplings produce a rugged terrain of compromise states—the Hopfield network introduced content-addressable memory, attractor dynamics, and a radical idea: that computation could be relaxation toward a minimum rather than the execution of instructions. Four decades later, its mathematics was rediscovered inside the attention mechanism of every
transformer, and the 2024 Nobel Prize in Physics made the lineage official. Yet on the day of his Nobel, the ninety-one-year-old physicist did not celebrate. He said he was “very unnerved” by a technology that works as a marvel while remaining incomprehensible—and he reached, without melodrama, for the comparison to nuclear fission. That refusal to