AI Wiki

Geoffrey Hinton

Artificial IntelligenceMachine Learning
11 min read
Updated
Suggest editHistoryTalk29 citationsFact-checked
RawGraph

Geoffrey Hinton (born December 6, 1947) is a British-Canadian computer scientist whose research has focused on neural networks, representation learning, and computational accounts of learning in the brain.[1][2] He is University Professor Emeritus at the University of Toronto and has served as chief scientific adviser to the Vector Institute since 2017.[1][4] Hinton shared the 2024 Nobel Prize in Physics with John J. Hopfield "for foundational discoveries and inventions that enable machine learning with artificial neural networks."[7] He also shared the 2018 ACM A.M. Turing Award, announced in 2019, with Yoshua Bengio and Yann LeCun for work on deep neural networks.[9]

News organizations often call Hinton the "godfather of AI," but this is a media nickname rather than a formal title or a claim of sole invention.[26][27] His best-known papers were collaborative. They include work on Boltzmann machines, backpropagation, the wake-sleep algorithm, deep belief networks, speech recognition, AlexNet, dropout, knowledge distillation, and capsule networks.[12][13][14][15][17][18][19][21][22]

Education and career

Hinton was born in London, United Kingdom. He completed a BA in experimental psychology at the University of Cambridge in 1970 and a PhD in artificial intelligence at the University of Edinburgh; the degree was awarded in 1978.[1][7] His curriculum vitae records research appointments at the University of Sussex, the University of California, San Diego, and the Medical Research Council Applied Psychology Unit in Cambridge before he joined Carnegie Mellon University's computer science faculty in 1982.[1]

Hinton moved to the University of Toronto in 1987. From 1998 through 2001, he was founding director of the Gatsby Computational Neuroscience Unit at University College London, which UCL says was established in July 1998.[1][3] He then returned to Toronto. His University of Toronto biography states that he directed the Canadian Institute for Advanced Research program on Neural Computation and Adaptive Perception from 2004 through 2013.[2] He became an emeritus professor at Toronto in 2014.[1]

Hinton operated a company called DNNresearch. In March 2013, Google acquired the company for an undisclosed amount and hired Hinton along with his University of Toronto graduate students Alex Krizhevsky and Ilya Sutskever; contemporary reporting does not support a precise purchase price.[5] Hinton worked half-time at Google from 2013 through 2023, first as a distinguished researcher and later as a vice president and engineering fellow.[1][2] He also became chief scientific adviser to the Vector Institute in January 2017.[1][4]

Hinton left Google in 2023 and said he did so to speak freely about AI risks, not to criticize Google.[6][27] Contemporary reporting described his concerns about misinformation, employment, malicious use, and loss of control.[27] As of the July 28, 2026 research cutoff for this article, he remained a University of Toronto professor emeritus and conducted AI-safety advocacy through the university's Schwartz Reisman Institute.[28][29]

Selected research contributions

The entries below describe what particular publications reported. They do not assign sole credit for fields with many contributors.

YearWorkHinton's documented role and evidence boundary
1985Boltzmann machine learningDavid Ackley, Hinton, and Terrence Sejnowski published a learning algorithm for stochastic networks with symmetric connections. The paper described how hidden units could learn internal representations from the difference between statistics measured in clamped and freely running phases.[12]
1986Learning representations by back-propagating errorsDavid Rumelhart, Hinton, and Ronald Williams described a procedure that repeatedly adjusted network weights to reduce the difference between actual and desired outputs, allowing hidden units to acquire useful features.[13] The paper documents Hinton's coauthorship; it does not support treating him as the method's sole inventor.
1995Wake-sleep algorithmHinton, Peter Dayan, Brendan Frey, and Radford Neal presented an unsupervised algorithm for layered stochastic networks. Its wake phase updated a generative model from data-driven hidden states, while its sleep phase updated a recognition model using samples from the generative model.[14]
2006Deep belief nets and deep autoencodersHinton, Simon Osindero, and Yee-Whye Teh proposed greedy layer-by-layer training for deep belief nets, followed by a fine-tuning procedure.[15] In a separate paper, Hinton and Ruslan Salakhutdinov reported an initialization method that let deep autoencoders learn low-dimensional codes under the experiments described in that study.[16]
2012Speech recognition and AlexNetA multi-laboratory review coauthored by Hinton summarized results in which deep neural networks improved acoustic modeling for speech recognition.[17] Krizhevsky, Sutskever, and Hinton then reported a 15.3% top-5 test error for their seven-network ensemble entry in the ILSVRC-2012 competition, compared with 26.2% for the second-best entry.[18] The result belongs to all three authors and does not make Hinton the inventor of convolutional networks.
2014DropoutNitish Srivastava, Hinton, Krizhevsky, Sutskever, and Salakhutdinov described training neural networks while randomly omitting units, then approximating an ensemble at test time. Their paper reported improvements on several supervised-learning tasks under its experimental settings.[19]
2015Knowledge distillationHinton, Oriol Vinyals, and Jeff Dean described training a smaller model with a larger model's softened output distribution as one way to transfer knowledge.[21] This supports attribution to the three-author paper, not a sole-inventor claim.
2017Capsule networksSara Sabour, Nicholas Frosst, and Hinton proposed dynamic routing between vector-valued capsules and reported results on MNIST and highly overlapping digits.[22] The paper was an experimental proposal; it did not establish capsules as a replacement for convolutional networks.
2022Forward-Forward algorithmHinton proposed replacing the forward and backward passes of backpropagation with two forward passes using positive and negative data. The paper explicitly presented preliminary results on small problems and called for further investigation.[23]

Boltzmann machines and the Nobel Prize

The 1985 Boltzmann-machine paper treated learning as a statistical process in a network of stochastic binary units.[12] The Nobel Committee's scientific background places that work after Hopfield's recurrent associative-memory model and explains how Hinton extended the statistical-physics connection to learning distributions over visible patterns with hidden nodes.[8] This chronology matters: the 2024 prize recognized Hopfield and Hinton for related but distinct contributions, while the original Boltzmann-machine learning paper also included Ackley and Sejnowski.[7][8][12]

Backpropagation and representation learning

The 1986 Nature paper showed how a multilayer network could learn internal representations by propagating output error information through the network.[13] Hinton's own biography describes him as "one of the researchers" who introduced backpropagation, wording that is more accurate than treating the algorithm as his individual invention.[2] His later work continued to investigate learned representations through generative models, deep belief nets, autoencoders, speech systems, and image classifiers.[14][15][16][17][18]

Deep learning at scale

AlexNet's competition result is often used as a marker in histories of modern deep learning. The paper describes a base network with 60 million parameters and 650,000 neurons trained across two graphics processors. For ILSVRC-2012, its Table 2 reports a 15.3% top-5 test error for an ensemble of seven convolutional networks, compared with 26.2% for the second-best entry.[18] An Associated Press account of the 2024 Nobel announcement called the competition a significant moment and quoted ImageNet creator Fei-Fei Li describing its historical importance.[26] Those later assessments are third-party characterizations, separate from the benchmark numbers in the paper itself.

Hinton also coauthored a 2015 review with LeCun and Bengio that surveyed representation learning with deep feedforward, convolutional, and recurrent networks.[20] The review documents a broad research program; it does not imply that any one of its authors invented every method it covers.

Awards and honors

This table lists selected major awards whose conferring organizations provide a clear citation. It is not a complete list.

AwardYearScope
ACM A.M. Turing Award2018Shared with Bengio and LeCun for conceptual and engineering breakthroughs associated with deep neural networks. ACM announced the award on March 27, 2019.[9]
Royal Medal2022The Royal Society cited Hinton's work on algorithms that learn distributed representations and their application to speech and vision.[10]
Nobel Prize in Physics2024Shared equally with Hopfield for foundational discoveries and inventions enabling machine learning with artificial neural networks.[7][8]
Queen Elizabeth Prize for Engineering2025Hinton was one of seven recipients. The prize body grouped Hinton with Bengio, Hopfield, and LeCun for the conceptual foundations of the neural-network approach, alongside separate hardware and benchmark-data contributions from the other recipients.[11]

Public statements about AI risk

Hinton's public warnings are his assessments, not established predictions. In his May 2023 CNN interview, he separated existing concerns, such as election manipulation and job displacement, from a longer-term possibility that more capable systems could escape human control.[6] The Guardian reported the same distinction and quoted his concern that generative systems could be used for misinformation by malicious actors.[27]

Hinton later coauthored "Managing extreme AI risks amid rapid progress," a 2024 Science policy forum article. The paper discusses large-scale social harms, malicious use, and possible loss of human control, while explicitly noting a lack of consensus about how such risks would arise or be managed.[24] It calls for technical safety research and governance mechanisms; it does not present an experimentally measured probability of catastrophe.

In an official Nobel interview recorded in December 2024 and published in 2026, Hinton again divided the issue into shorter-term misuse and longer-term control risks. He stressed uncertainty, saying that confident claims of either safety or inevitable takeover went beyond what was known, and argued for more basic research on control.[25] The University of Toronto established support for his safety advocacy through the Schwartz Reisman Institute by January 2026.[29]

Reception and attribution

Award bodies and independent reporting characterize Hinton as an influential pioneer, but they divide credit among collaborators. ACM's award went jointly to Bengio, Hinton, and LeCun.[9] The Nobel Committee linked Hinton's Boltzmann-machine work to Hopfield's earlier network and the wider history of neural networks.[8] The Queen Elizabeth Prize recognized seven contributors spanning neural-network methods, computer hardware, and ImageNet.[11]

Independent coverage also records disagreement about Hinton's risk assessments. In its 2024 Nobel report, the Associated Press quoted former student Nicholas Frosst agreeing that the risks should be discussed while disputing Hinton's timescale and the danger posed by then-current systems.[26] This disagreement does not alter Hinton's research record, but it is important context when reporting his forward-looking claims.

See also

References

  1. ^Geoffrey E. Hinton. "Curriculum Vitae." January 6, 2025. cs.toronto.edu/...shortcv.pdf
  2. ^Geoffrey E. Hinton. "Geoffrey E. Hinton's Biographical Sketch." University of Toronto. cs.utoronto.ca/...bio
  3. ^University College London. "About: Gatsby Computational Neuroscience Unit." ucl.ac.uk/...about
  4. ^Vector Institute. "Vector Institute's Chief Scientific Advisor, Dr. Geoffrey Hinton, receives ACM A.M. Turing Award alongside Dr. Yoshua Bengio and Dr. Yann LeCun." March 27, 2019. vectorinstitute.ai/...hua-bengio-and-dr-yann-lecun
  5. ^Robert McMillan. "Google Hires Brains that Helped Supercharge Machine Learning." *Wired*, March 13, 2013. wired.com/...google-hinton
  6. ^CNN. "AI Pioneer Addresses Dangers of Technology." Transcript, May 3, 2023. transcripts.cnn.com/...05
  7. ^Nobel Prize Outreach. "Geoffrey Hinton: Facts." Nobel Prize in Physics 2024. nobelprize.org/...hinton
  8. ^Nobel Committee for Physics. "Scientific Background to the Nobel Prize in Physics 2024." October 8, 2024. nobelprize.org/...advanced-physicsprize2024-3.pdf
  9. ^Association for Computing Machinery. "Fathers of the Deep Learning Revolution Receive ACM A.M. Turing Award." March 27, 2019. awards.acm.org/...turing-award-2018.pdf
  10. ^Royal Society. "The Royal Society announces this year's medal and award winners." August 23, 2022. royalsociety.org/...medals-and-awards-2022
  11. ^Queen Elizabeth Prize for Engineering. "2025 QEPrize Winners: Modern Machine Learning." qeprize.org/...modern-machine-learning
  12. ^David H. Ackley, Geoffrey E. Hinton, and Terrence J. Sejnowski. "A Learning Algorithm for Boltzmann Machines." *Cognitive Science* 9, no. 1 (1985): 147-169. cs.toronto.edu/...cogscibm.pdf
  13. ^David E. Rumelhart, Geoffrey E. Hinton, and Ronald J. Williams. "Learning representations by back-propagating errors." *Nature* 323 (1986): 533-536. nature.com/...323533a0
  14. ^Geoffrey E. Hinton, Peter Dayan, Brendan J. Frey, and Radford M. Neal. "The wake-sleep algorithm for unsupervised neural networks." *Science* 268, no. 5214 (1995): 1158-1161. cs.toronto.edu/...ws.pdf
  15. ^Geoffrey E. Hinton, Simon Osindero, and Yee-Whye Teh. "A Fast Learning Algorithm for Deep Belief Nets." *Neural Computation* 18, no. 7 (2006): 1527-1554. cs.toronto.edu/...fastnc.pdf
  16. ^Geoffrey E. Hinton and Ruslan R. Salakhutdinov. "Reducing the Dimensionality of Data with Neural Networks." *Science* 313, no. 5786 (2006): 504-507. pubmed.ncbi.nlm.nih.gov/16873662
  17. ^Geoffrey Hinton et al. "Deep Neural Networks for Acoustic Modeling in Speech Recognition." *IEEE Signal Processing Magazine* 29, no. 6 (2012): 82-97. research.google/...-modeling-in-speech-recognition
  18. ^Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. "ImageNet Classification with Deep Convolutional Neural Networks." *Advances in Neural Information Processing Systems* 25 (2012). proceedings.neurips.cc/...436e924a68c45b-Paper.pdf
  19. ^Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. "Dropout: A Simple Way to Prevent Neural Networks from Overfitting." *Journal of Machine Learning Research* 15 (2014): 1929-1958. jmlr.org/...srivastava14a
  20. ^Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. "Deep learning." *Nature* 521 (2015): 436-444. nature.com/...nature14539
  21. ^Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. "Distilling the Knowledge in a Neural Network." arXiv:1503.02531, 2015. arxiv.org/...1503.02531
  22. ^Sara Sabour, Nicholas Frosst, and Geoffrey E. Hinton. "Dynamic Routing Between Capsules." arXiv:1710.09829, 2017. arxiv.org/...1710.09829
  23. ^Geoffrey Hinton. "The Forward-Forward Algorithm: Some Preliminary Investigations." arXiv:2212.13345, 2022. arxiv.org/...2212.13345
  24. ^Yoshua Bengio et al. "Managing extreme AI risks amid rapid progress." *Science* 384, no. 6698 (2024): 842-845. arxiv.org/...2310.17688
  25. ^Nobel Prize Outreach. "Transcript from an interview with Geoffrey Hinton." Interview recorded December 6, 2024; published June 3, 2026. nobelprize.org/...1925103-interview-transcript
  26. ^Associated Press. "Pioneers in artificial intelligence win the Nobel Prize in physics." October 8, 2024. ap.org/...elligence-win-the-nobel-prize-in-physics
  27. ^Josh Taylor and Alex Hern. "'Godfather of AI' Geoffrey Hinton quits Google and warns over dangers of misinformation." *The Guardian*, May 2, 2023. theguardian.com/...rns-dangers-of-machine-learning
  28. ^University of Toronto. "In his words: Geoffrey Hinton reflects on his Nobel Prize win." October 10, 2024. utoronto.ca/...hinton-reflects-his-nobel-prize-win
  29. ^University of Toronto. "What happens when AI is smarter than us? Gift supports Geoffrey Hinton's global AI safety mission." January 16, 2026. utoronto.ca/...y-hinton-s-global-ai-safety-mission

Improve this article

Add missing citations, update stale details, or suggest a clearer explanation. Every suggestion is reviewed for sourcing before it goes live.

7 revisions · v8 · 2,292 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent 2026-07-28 fact-check: 29 explicit references, 74 citation calls, eight canonical internal targets, and 20 high-risk claim groups checked. Root inspected 28 production renders plus selected source renders. Corrected AlexNet's base-network versus seven-CNN ensemble scope, bounded backpropagation attribution to the cited paper, replaced a mutable Royal Medal source with the direct 2022 announcement, followed the QEPrize body's division of credit, removed an unsupported expanded middle name, and retained the contemporary record that DNNresearch's purchase price was undisclosed.

Cite this page: AI Wiki. "Geoffrey Hinton." aiwiki.ai, updated 30 Jul 2026, fact-checked 30 Jul 2026. CC BY 4.0. https://aiwiki.ai/wiki/geoffrey_hinton

Suggest edit