PERSON
David Rumelhart
The quietest man in cognitive science who made backpropagation practical, co-authored the connectionist manifesto that became the blueprint of modern AI, and—in a symmetry almost too apt to state—spent his final years watching his own distributed representations dissolve under a dementia that attacked exactly the faculties he had spent his life modeling.
David Everett Rumelhart is the reason every deep neural network alive can learn anything at all. The algorithm he made practical—
backpropagation—trains every deployed model, every image classifier, every language system, every game-playing agent that learned its skill. Together with James McClelland he edited the two-volume
Parallel Distributed Processing (1986), the connectionist manifesto that argued the mind was not a digital computer running a program but a vast network of simple units whose intelligence emerged from the pattern of their interaction rather than from any rule written anywhere. That claim was dismissed as a romantic detour when Rumelhart proposed it; it became, four decades later, the literal engineering blueprint of modern AI. He spent his career building models of how the mind learns—how knowledge lives not in any single place but spread across countless connections, how such a system fails not all at