1947–, Computer scientist; "godfather of deep learning"
Also known as: Geoffrey E. Hinton, Geoff Hinton
Geoffrey Everest Hinton is a British-Canadian computer scientist who, more than any other single individual, founded modern deep learning. As a co-author of the 1986 paper Learning Representations by Back-Propagating Errors with David Rumelhart and Ronald Williams, he helped popularise the backpropagation algorithm that made multi-layer neural networks practically trainable. The paper relaunched neural-network research after the cooling that followed Minsky and Papert's Perceptrons (1969).
Over the next four decades Hinton produced or co-produced nearly every important neural-network method. He co-invented the Boltzmann machine (with Terry Sejnowski, 1985) and the restricted Boltzmann machine; pioneered distributed representations; co-developed the wake–sleep algorithm for unsupervised learning; introduced dropout (with his students Krizhevsky, Sutskever and Salakhutdinov, 2012); supervised the AlexNet result (Krizhevsky and Sutskever, 2012) that triggered the modern deep-learning revolution; co-developed the t-SNE dimensionality-reduction algorithm (with Laurens van der Maaten, 2008); and proposed capsule networks (2017–).
Hinton's career has alternated between Carnegie Mellon, the University of Toronto (his long-time home), and Google Brain (2013–2023). His students and postdocs include Yann LeCun, Yoshua Bengio (jointly), Ilya Sutskever, Alex Krizhevsky, Ruslan Salakhutdinov, Brendan Frey, Nitish Srivastava, Roland Memisevic, Vinod Nair and many others, the leadership of the modern neural-network community is, in large part, his academic descent.
He shares the 2018 Turing Award with Yann LeCun and Yoshua Bengio for their joint work on deep learning, and the 2024 Nobel Prize in Physics with John Hopfield. In 2023 he resigned from Google to speak more freely about AI safety, becoming one of the most prominent voices warning about advanced AI risks. On his father's side his great-great-grandfather is George Boole, and his great-grandfather is Charles Howard Hinton, a mathematician known for work on higher-dimensional geometry.
Video
Related people: David Rumelhart, Yann LeCun, Yoshua Bengio, John Hopfield, George Boole
Works cited in this book:
- Learning representations by back-propagating errors (1986) (with David E. Rumelhart, Ronald J. Williams)
- Visualizing Data using t-SNE (2008) (with Laurens van der Maaten)
- Rectified Linear Units Improve Restricted Boltzmann Machines (2010) (with Vinod Nair)
- ImageNet Classification with Deep Convolutional Neural Networks (2012) (with Alex Krizhevsky, Ilya Sutskever)
- Dropout: A Simple Way to Prevent Neural Networks from Overfitting (2014) (with Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, Ruslan Salakhutdinov)
- Distilling the Knowledge in a Neural Network (2015) (with Oriol Vinyals, Jeff Dean)
- Layer Normalization (2016) (with Jimmy Lei Ba, Jamie Ryan Kiros)
- A Simple Framework for Contrastive Learning of Visual Representations (2020) (with Ting Chen, Simon Kornblith, Mohammad Norouzi)
Discussed in:
- Chapter 1: What Is AI?, A Brief History of AI
- Chapter 9: Neural Networks, Neural Networks