The best-documented honors for modern AI pioneers are two prizes. The Association for Computing Machinery named Yoshua Bengio, Geoffrey Hinton, and Yann LeCun recipients of the 2018 Turing Award on March 27, 2019, and the Royal Swedish Academy of Sciences gave the 2024 Nobel Prize in Physics to John Hopfield and Geoffrey Hinton on October 8, 2024. Both citations concern artificial neural networks, the approach behind most current AI systems, and both leave out many other contributors.
Key takeaways
- The 2018 Turing Award, carrying a USD 1 million prize, recognized Bengio, Hinton, and LeCun for conceptual and engineering work on deep neural networks.
- The 2024 Nobel Prize in Physics, worth 11 million Swedish kronor shared equally, went to Hopfield (Princeton University) and Hinton (University of Toronto).
- Hinton's 1986 backpropagation paper in Nature had three authors, so the prizes name a subset of the people involved.
- Two of the names are tied to Canadian institutions: Hinton to the University of Toronto and Bengio to the University of Montreal and Mila.
- A prize citation is a statement by a committee. It is not a complete history of who invented what.
What the Turing Award committee credited
ACM announced the 2018 award on March 27, 2019, for conceptual and engineering breakthroughs that made deep neural networks a central part of computing. At the time Bengio was a professor at the University of Montreal and scientific director of Mila, Hinton held positions at Google, the Vector Institute, and the University of Toronto, and LeCun was a professor at New York University and chief AI scientist at Facebook. ACM said the three developed conceptual foundations, identified surprising phenomena through experiments, and made engineering advances that showed the practical value of deep networks.
The announcement also gave a reason for the timing. Neural networks had been introduced in the 1980s, but by the early 2000s only a small group stayed committed to them, and their efforts met skepticism at first. ACM linked the later success to two outside factors, powerful graphics processing units and access to massive datasets. The award honored a 30-year effort, not a single discovery.
What each person is credited with
The table follows the technical accomplishments ACM listed in its announcement, with the Nobel committee's wording for Hopfield. Dates are those given in the sources.
| Person | Credited contribution | Year given by the source |
|---|---|---|
| Geoffrey Hinton | Boltzmann machines, with Terrence Sejnowski; one of the first networks able to learn internal representations | 1983 |
| Geoffrey Hinton | Backpropagation shown to let networks discover their own internal representations, with David Rumelhart and Ronald Williams | 1986 |
| Geoffrey Hinton | Improved convolutional networks with students Alex Krizhevsky and Ilya Sutskever, nearly halving the ImageNet object recognition error rate | 2012 |
| Yann LeCun | Convolutional neural networks, first trained on images of handwritten digits | 1980s |
| Yoshua Bengio | Neural networks combined with probabilistic sequence models, used in a system for reading handwritten checks | 1990s |
| Yoshua Bengio | "A Neural Probabilistic Language Model," which introduced high-dimension word embeddings | 2000 |
| John J. Hopfield | An associative memory network that can store and reconstruct patterns such as images | 1980s onward |
Why a physics prize went to machine learning
The Nobel committee's citation reads "for foundational discoveries and inventions that enable machine learning with artificial neural networks." The press release explains the link to physics. The Hopfield network is described using the physics of atomic spin, with the saved images stored as low-energy states, so that a distorted or incomplete image can be reconstructed by working through the nodes. Hinton used tools from statistical physics to build the Boltzmann machine on that foundation. The machine learns characteristic elements in a type of data and can classify images or produce new examples of the pattern it was trained on.
The release says the laureates' work on artificial neural networks dates from the 1980s onward. It quotes Ellen Moons, chair of the Nobel Committee for Physics, saying that physicists use artificial neural networks in a vast range of areas, such as developing new materials with specific properties. The prize therefore recognizes the method and its use in physics, and the committee does not claim that it settles who built today's chatbots.
Where the Canadian connection sits
Hinton's affiliation in the Nobel announcement is the University of Toronto. Bengio's in the ACM announcement is the University of Montreal and Mila, described there as Quebec's Artificial Intelligence Institute. ACM also noted that the three continued to explore machine learning, neuroscience, and cognitive science through the Learning in Machines and Brains program of CIFAR, formerly the Canadian Institute for Advanced Research. LeCun's own early work included a period at the University of Toronto in the late 1980s, according to the same announcement. Institutions are one part of the story, and the announcements do not say which funding decisions mattered most.
What these prizes do not tell
A prize names a few people because it can name only a few. The 1986 backpropagation paper, published in Nature on October 9, 1986, lists Rumelhart, Hinton, and Williams as authors. ACM itself describes Hinton, LeCun, and Bengio as part of a small group that stayed with neural networks, which implies others were involved. Computer science awards and physics awards also follow different rules and different nomination pools, so a missing name is not evidence of a missing contribution.
The prizes also say nothing about whether today's systems are safe or useful. The Turing citation covers technical contributions, and the Nobel citation is limited to foundations. Earlier origin questions, such as Turing's 1950 paper and the Dartmouth proposal of 1955, are a separate chapter of the field's history.
Questions to ask about any list of AI pioneers
- Is each credit tied to a dated paper, award citation, or patent?
- Does the list say which part of a result a person produced and which part came from co-authors?
- Are prizes treated as evidence of priority, or only as a committee's judgment?
- Does the list cover people outside the best-known universities and companies?
- Does it separate the pioneers of a method from the builders of a particular product?
FAQ
Who are called the godfathers of deep learning?
The phrase is informal. ACM's 2019 announcement used the more formal wording that Bengio, Hinton, and LeCun developed the conceptual foundations of deep learning, and the press release was headlined as honoring the fathers of the deep learning revolution.
Did Geoffrey Hinton invent backpropagation alone?
No. The 1986 Nature paper has three authors, and ACM says that LeCun proposed an early version of the algorithm and gave a derivation of it. The paper's contribution, as described by ACM, was showing that backpropagation lets networks learn useful internal representations.
Why did a physicist share a prize about machine learning?
Hopfield's network uses the mathematics of atomic spin systems. The Nobel committee judged that the physics-based foundations of neural networks fall within the physics prize.
Are the Turing Award and the Nobel Prize the same kind of honor?
No. The Turing Award is given by a professional society for contributions to computing, and the Nobel Prize in Physics is given by the Royal Swedish Academy of Sciences. The amounts, committees, and criteria differ.
Related reading
- Early figures in the origins of AI
- The minds behind artificial intelligence
- Why artificial intelligence was created
- Main topics and applications of AI
More in AI Basics.
Last verified: October 2026
Sources
- Fathers of the Deep Learning Revolution Receive ACM A.M. Turing Award, ACM press release, March 27, 2019
- Press release: The Nobel Prize in Physics 2024, Royal Swedish Academy of Sciences, October 8, 2024
- Rumelhart DE, Hinton GE, Williams RJ, Learning representations by back-propagating errors, Nature, volume 323, October 9, 1986