The Connectionist Revival

Geoffrey Hinton

Known as a godfather of deep learning — co-authored backpropagation in 1986, co-invented the Boltzmann machine and dropout, and co-authored AlexNet, the 2012 result that restarted the field.

Geoffrey Hinton (b. 1947) spent most of his career as a minority voice insisting neural networks would eventually work, through two AI winters that suggested otherwise. With David Rumelhart and Ronald Williams, he co-authored the 1986 paper that popularized backpropagation. Across the following decades he kept building the toolkit deep learning would eventually run on: Boltzmann machines, restricted Boltzmann machines used for pretraining before better initialization made that unnecessary, and dropout — randomly disabling neurons during training, still one of the most widely used regularization techniques.

The result that changed everything came in 2012. Hinton, with students Ilya Sutskever and Alex Krizhevsky, built AlexNet, a CNN that won the ImageNet competition by a margin so large — roughly halving the error rate of the next-best entry — that it ended a long-running argument about whether deep networks could outperform hand-engineered computer vision features. AlexNet is generally treated as the moment deep learning went from a promising niche to the dominant paradigm in machine learning.

Hinton, Yann LeCun, and Yoshua Bengio shared the 2018 Turing Award — computing's equivalent of the Nobel Prize — for their contributions to deep learning, and the trio is commonly called the "godfathers of deep learning." In 2023, Hinton left Google specifically to speak more freely about AI safety risks he believes the field is not taking seriously enough.

See also: David Rumelhart, Yann LeCun, Yoshua Bengio, Ilya Sutskever

Learn more: Neural Networks & Backprop · Computer Vision · Wikipedia: Geoffrey Hinton

Mentioned in

Lessons where this comes up in context.