Generative Art Timeline · Chapter 6: Personal Computer Era (1980s)

1986

Backpropagation Works

Publication · North America

David Rumelhart at the International Joint Conference on Neural Networks, Seattle, 1991.
David Rumelhart at the International Joint Conference on Neural Networks, Seattle, 1991. Photo: Rolf Kickuth, CC BY-SA 4.0, via Wikimedia Commons.

Their Nature paper showed how to train the multilayer networks Minsky and Papert had doubted. Errors were passed backwards through the network so every connection learned its share of the blame. The idea itself wasn’t new, having been sketched in the 1970s and ignored, but now backpropagation worked. After seventeen years, the learning camp was back.

Text by Peter Bauman. Generative Art Timeline, Le Random.

Sources

Movements

artificial intelligence

Filed under

Ideas
ai history
Technologies
artificial intelligence, backpropagation, neural networks
Organisations
UC San Diego
Other subjects
Carnegie Mellon University, David Rumelhart, Geoffrey Hinton, Ronald Williams

Related moments

Le Random editorials

Explore all Le Random editorials and artist interviews

Le Random podcast