"Hopfield’s seminar helped the Connection Machine team discover how its architecture could support neural-network computation."
Hillis describes Feynman translating Hopfield's neural-network approach into computations suitable for the Connection Machine's processors. The episode demonstrates how inviting outside expertise can expose useful applications of an unfamiliar technology and generate new technical work. Hillis's account appeared in 1989 and was republished by Long Now in 2017.
View Source