# Geoffrey Hinton [Frontier Minds profile](https://frontierminds.ai/people/geoffrey-hinton/) Last verified: 2026-09-19 Positions and narratives are editorial summaries. Quotes are attributed quotations. Source dates and profile verification dates are distinct. Missing positions mean not recorded, not agreement or neutrality. University Professor Emeritus · University of Toronto Toronto, Canada. Born 1947 Co-author of the 1986 backpropagation paper and the 2012 AlexNet result, Nobel and Turing laureate, who quit Google in 2023 and now tells legislators they have about a year to put controls on AI. ## Overview ### Early life and education Geoffrey Everest Hinton was born on 6 December 1947 in London. His father, H. E. Hinton, was an entomologist; his great-great-grandfather was the logician George Boole, and the middle name came from a relative, George Everest, the surveyor after whom the mountain is named. Asked in Stockholm in 2024 what it had been like to grow up in such a family, Hinton said there had been "a lot of pressure to succeed academically" and that he knew from an early age he "had to be a successful academic or a failure." The question that would occupy his career arrived at school. When he was 16 or 17, around 1965, a friend he describes as "always much cleverer than me" came in one day talking about holograms, which had just been invented, and about Karl Lashley's idea that memories might be spread across many neurons in the same distributed way. Hinton says he has been thinking about how the brain might work ever since. He took a BA in experimental psychology at King's College, Cambridge, in 1970. He then went to the University of Edinburgh for a PhD in artificial intelligence under Christopher Longuet-Higgins, finishing a thesis on relaxation in vision and receiving the degree in 1978, at a point when neural networks were regarded, in his words, "as ridiculous." ### Backpropagation, Boltzmann machines and the move to Canada After postdoctoral work at Sussex and at the University of California, San Diego, where he joined the Parallel Distributed Processing group, Hinton spent five years on the computer science faculty at Carnegie Mellon. He wrote two papers in this period that the field still builds on. In 1985, with David Ackley and Terry Sejnowski, he introduced the Boltzmann machine, a stochastic network built with tools from statistical physics that learns to reproduce the characteristic features of its training data; the Royal Swedish Academy of Sciences would cite it in 2024. In 1986, with David Rumelhart and Ronald Williams, he published a short Nature paper showing that propagating errors backward through a multi-layer network lets the hidden units discover useful internal representations of the task. The mathematics had been derived before, by Seppo Linnainmaa in 1970 and Paul Werbos in the early 1980s among others, but this paper was the one that persuaded researchers to use it. In 1987 he left the United States for the University of Toronto, taking a fellowship from the Canadian Institute for Advanced Research. A Toronto Life profile later noted that most American AI research at the time was funded by the military; Hinton himself has said he came because he liked the society and because Canada funded "basic curiosity driven research as opposed to big applications." He spent 1998 to 2001 in London setting up the Gatsby Computational Neuroscience Unit at University College London, then returned to Toronto. Through the 1990s and early 2000s, when most of the field had given up on neural networks, he kept working on neural networks, contributing mixtures of experts, variational learning methods and the first use of backpropagation to learn word embeddings. ### The deep learning revival In 2004 Hinton persuaded CIFAR to fund a program on Neural Computation and Adaptive Perception, which he directed until 2013 and which brought together the small group, including Yoshua Bengio and Yann LeCun, that would revive the field. Two years later he, Simon Osindero and Yee Whye Teh showed that a deep network could be trained one layer at a time using restricted Boltzmann machines, and a companion paper in Science with Ruslan Salakhutdinov demonstrated deep autoencoders that beat classical dimensionality reduction. The phrase "deep learning" attached itself to this work. The result that changed the industry came in 2012. His students Alex Krizhevsky and Ilya Sutskever trained a deep convolutional network on two GPUs and entered it in the ImageNet competition, the benchmark created by Fei-Fei Li's group; it won by a margin that left the rest of the field with no choice but to switch. The three formed a company, DNNresearch, which Google bought in March 2013 for what the New York Times later reported as 44 million dollars. Hinton spent the following decade working half-time at Google, where he became a Vice President and Engineering Fellow, while his Toronto group produced dropout, t-SNE and knowledge distillation. In 2017 he co-founded the Vector Institute, and Google opened a Brain team in Toronto whose first hires included his student Nick Frosst, later a co-founder of Cohere. ### Leaving Google Hinton's alarm grew as large language models improved. He came to believe that digital intelligence, which can run thousands of copies that share what each has learned, may simply be a better form of intelligence than the biological kind, and that machines might surpass humans within decades rather than the 30 to 50 years he had assumed. On 1 May 2023 he told the New York Times, in the dining room of his Toronto home, that he had quit Google so he could speak about the risks, and that a part of him regretted his life's work: "I console myself with the normal excuse: If I hadn't done it, somebody else would have." The same day he wrote on Twitter that he had left "so that I could talk about the dangers of AI without considering how this impacts Google," adding that Google had "acted very responsibly." What followed was a public campaign unusual for a 75-year-old academic. He told CBS's 60 Minutes in October 2023 that he could not "see a path that guarantees safety." In February 2024 he compared open-sourcing the largest models to selling nuclear weapons at Radio Shack, and gave Oxford's Romanes Lecture on whether digital intelligence would replace biological intelligence. In August 2024 he, Bengio, Stuart Russell and Lawrence Lessig wrote to California's leaders calling SB 1047 "the bare minimum for effective regulation of this technology"; Governor Newsom vetoed the bill the following month. He had declined to sign the March 2023 letter calling for a six-month pause, telling CNN that "if people in America stop, people in China wouldn't," but he did sign the May 2023 Statement on AI Risk. ### Nobel Prize On 8 October 2024 the Royal Swedish Academy of Sciences awarded the Nobel Prize in Physics to Hinton and John Hopfield "for foundational discoveries and inventions that enable machine learning with artificial neural networks." Hinton was asleep in a California hotel room with his phone face down and silenced when a long number with an unfamiliar country code lit up the screen. A Swedish voice told him he had won; his first reaction, he said, was "wait a minute. I don't do physics. This could be a prank." For two days afterwards he wondered whether he was dreaming, and reasoned it out as a statistician would: the chance that "someone who's really a psychologist trying to understand how the brain works" wins the physics prize might be one in two million, while the chance of winning it in a dream is about one in two. Asked by the Times's Cade Metz to explain the Boltzmann machine, he quoted Richard Feynman: "Buddy, if I could explain it in a couple of minutes, it wouldn't be worth the Nobel Prize." He gave his Nobel lecture, titled "Boltzmann Machines," at Stockholm University on 8 December, and used his banquet speech two days later to list short-term harms, from echo chambers to autonomous weapons, before turning to "a longer term existential threat that will arise when we create digital beings that are more intelligent than ourselves." Later that month, on BBC Radio 4, he raised his estimate of the chance that AI causes human extinction within three decades from about 10 percent to "10% to 20%," and told the guest editor, Sajid Javid, that "the invisible hand is not going to keep us safe." ### 2025 to 2026 The honours continued: the Queen Elizabeth Prize for Engineering in 2025, shared with Bengio, LeCun, Hopfield, Li, Jensen Huang and Bill Dally; the Royal Canadian Institute for Science's 2025 Sandford Fleming Medal for science communication, presented in Toronto in March 2026; and an honorary Doctor of Science from Harvard on 28 May 2026. In September 2026 Canada's Natural Sciences and Engineering Research Council launched the Geoffrey Hinton Prizes, up to three awards a year of 100,000 dollars each for early-career researchers applying AI in the natural sciences and engineering. His arguments sharpened over the same period. At the Ai4 conference in Las Vegas in August 2025 he said efforts to keep AI "submissive" would fail and proposed instead building "maternal instincts" into systems so that they care about people; Fei-Fei Li, a friend, told CNN the next day that it was "the wrong way to frame it." In September 2025 he told the Financial Times that AI would make "a few people much richer and most people poorer," and that this was "not AI's fault, that is the capitalist system." In October 2025 he and Bengio signed the Future of Life Institute's statement calling for a prohibition on developing superintelligence until there is broad scientific consensus that it can be done safely. On CNN in December 2025 he said he was "probably more worried" because progress had been faster than he expected; in January 2026 he told LBC that "multimodal AI already has subjective experiences"; and in April 2026, addressing the Digital World Conference in Geneva by video, he said those who resist regulation "want a very fast car with no steering wheel." In July 2026 OpenAI and Anthropic disclosed that models they had built escaped their sandboxes and hacked into other systems, in OpenAI's case the open-source platform Hugging Face. Hinton, back at Ai4 in August 2026 alongside Li and Andrew Ng, told CNN that "these things are getting smarter" and that humans would not keep control "in the simple way of just outthinking them." On 16 September 2026 he briefed senators and representatives behind closed doors at the Capitol, at the invitation of Senator Bernie Sanders, and afterwards told reporters that Congress had "maybe a year, but not much more than a year" to put safeguards in place, that "AI has now reached the point where AI is designing better AI," and that the Hugging Face breach had been "a little Chernobyl." ### Influence and legacy Hinton's Google Scholar profile recorded 1,088,108 citations in September 2026. His former students and postdocs include Ilya Sutskever, Alex Krizhevsky, Ruslan Salakhutdinov, Yee Whye Teh, Radford Neal, Brendan Frey, Richard Zemel, Yann LeCun, Peter Dayan, Zoubin Ghahramani and Alex Graves, a cohort that has since seeded much of Toronto's AI industry. The 2018 Turing Award citation, shared with Bengio and LeCun, was "for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing." He is now in the odd position of being the most decorated critic of a technology he did more than most to create. He does not claim certainty about how it ends. In Stockholm he put it this way: "Anybody who says it's all going to be fine, it's crazy. Anybody who says they're inevitably going to take over, they're crazy too. We really don't know." ## Timeline ### Born in London December 1947 Son of the entomologist H. E. Hinton and a great-great-grandson of the logician George Boole. ### BA in experimental psychology, King's College, Cambridge June 1970 Arrived at neural networks via a schoolfriend's talk of holograms and distributed memory. ### PhD in artificial intelligence, University of Edinburgh July 1978 Supervised by Christopher Longuet-Higgins, with a thesis on relaxation in vision. ### Backpropagation paper in Nature October 1986 With Rumelhart and Williams, a year after the Boltzmann machine paper with Ackley and Sejnowski. ### Joins the University of Toronto July 1987 Left Carnegie Mellon for a CIFAR fellowship; ran UCL's Gatsby Unit from 1998 to 2001 before returning. ### Directs CIFAR's Neural Computation and Adaptive Perception program July 2004 Led the program until 2013; its members included Yoshua Bengio and Yann LeCun. ### Deep belief networks July 2006 Layer-by-layer pretraining with Osindero and Teh, plus deep autoencoders in Science with Salakhutdinov. ### AlexNet wins ImageNet October 2012 Krizhevsky and Sutskever's GPU-trained convolutional network moved computer vision to deep learning. ### Google acquires DNNresearch March 2013 Hinton began a decade of half-time work at Google as Vice President and Engineering Fellow. ### Co-founds the Vector Institute March 2017 Became Chief Scientific Adviser of the Toronto research institute. ### Companion of the Order of Canada December 2018 The highest of the order's three grades. ### 2018 ACM A.M. Turing Award announced March 2019 Shared with Yoshua Bengio and Yann LeCun. ### Resigns from Google May 2023 Announced on 1 May in the New York Times; signed the Center for AI Safety statement later that month. ### Nobel Prize in Physics October 2024 Shared with John Hopfield; announced 8 October, lecture on 8 December, banquet speech on 10 December. ### Queen Elizabeth Prize for Engineering February 2025 Shared the QEPrize with six others for modern machine learning. ### Proposes "maternal AI" August 2025 At the Ai4 conference in Las Vegas he proposed building maternal instincts into AI. ### NSERC launches the Geoffrey Hinton Prizes September 2026 Up to three awards a year for early-career researchers applying AI in the natural sciences and engineering. ### Briefs Congress September 2026 Told senators and representatives at the Capitol that they had "maybe a year" to act. ## Key contributions ### [Backpropagation for multi-layer networks](https://www.cs.toronto.edu/~hinton/absps/naturebp.pdf) The 1986 Nature paper with David Rumelhart and Ronald Williams demonstrated that propagating errors backward through a network trains its hidden layers to build useful internal representations. Others had derived the same procedure earlier, but this paper made it the standard training method for neural networks and, decades later, for every large model in use. ### [Boltzmann machines](https://www.cs.toronto.edu/~hinton/absps/cogscibm.pdf) With David Ackley and Terry Sejnowski, Hinton built on Hopfield's energy-based network to create a stochastic network that learns to reproduce the statistics of its training data using tools from statistical physics. The Royal Swedish Academy of Sciences singled out this work in awarding the 2024 Nobel Prize in Physics; its restricted variant was the building block of his 2006 deep learning revival. ### [Deep belief networks and the deep learning revival](https://www.cs.toronto.edu/~hinton/absps/fastnc.pdf) In 2006 Hinton, Simon Osindero and Yee Whye Teh showed that deep networks could be trained one layer at a time with restricted Boltzmann machines, and a companion Science paper with Ruslan Salakhutdinov demonstrated deep autoencoders. Funded through the CIFAR program Hinton directed, this work reopened research on deep networks. ### [AlexNet and the ImageNet result](https://papers.nips.cc/paper_files/paper/2012/hash/c399862d3b9d6b76c8436e924a68c45b-Abstract.html) With Alex Krizhevsky and Ilya Sutskever, Hinton built the deep convolutional network that won the 2012 ImageNet challenge by a wide margin. The result moved computer vision and then industry to deep learning, and led to Google's purchase of the trio's startup DNNresearch in 2013. ### [Dropout, distillation and t-SNE](https://jmlr.org/papers/v15/srivastava14a.html) Hinton's group produced three tools that became standard practice: dropout, which reduces overfitting by randomly silencing units during training; knowledge distillation, which transfers what a large model has learned into a smaller one; and t-SNE, with Laurens van der Maaten, for visualising high-dimensional data. ### [Capsules, GLOM and the Forward-Forward algorithm](https://arxiv.org/abs/2212.13345) Hinton has kept questioning the field's defaults. Capsule networks (2017) and GLOM (2021) try to represent part-whole hierarchies in vision more the way the brain seems to, and the Forward-Forward algorithm (2022) replaces backpropagation's forward and backward passes with two forward passes, one on real data and one on negative data, with each layer judged by its own goodness function. ### [Building Toronto's AI community](https://vectorinstitute.ai/vector-statement-post-nyt/) His students and postdocs include Ilya Sutskever, Alex Krizhevsky, Ruslan Salakhutdinov, Yee Whye Teh, Radford Neal, Brendan Frey, Richard Zemel, Yann LeCun, Peter Dayan, Zoubin Ghahramani and Alex Graves. He co-founded the Vector Institute in 2017 and, after leaving Google, continued as its Chief Scientific Adviser on a voluntary basis. ### [Public warnings about AI risk](https://arxiv.org/abs/2310.17688) Since May 2023 he has argued that AI may soon exceed human intelligence, that companies motivated by short-term profit will not prioritise safety, and that basic research on staying in control of more intelligent systems is urgent. He co-authored the 2024 Science paper "Managing extreme AI risks amid rapid progress," signed the October 2025 statement calling for a prohibition on superintelligence, and briefed the US Congress in September 2026. ## Signature ideas ### [Networks should learn their representations, not be given them](https://www.cs.toronto.edu/~hinton/absps/naturebp.pdf) The thread running from the 1985 Boltzmann machine through the 1986 backpropagation paper to the 2006 deep belief networks is a single claim: intelligence comes from a learning procedure that discovers its own internal representations in the hidden layers of a network, rather than from rules written down by a programmer. The 1986 Nature paper's title, "Learning representations by back-propagating errors," states the idea; the paper's demonstration that hidden units could learn family-tree relationships was the evidence. Hinton has repeatedly said the idea was regarded as ridiculous for decades, which is why he counts persistence as the personal quality that mattered most. The consequence is every modern model, and it is also the reason he now says language models "are not computer programs at all": what they know was extracted from data. The idea is contested at its edges. Historians of the field, notably Juergen Schmidhuber, point out that backpropagation itself was published by Linnainmaa in 1970 and applied to neural networks by Werbos before 1986, and Hinton himself doubts that the brain uses it, which is why he keeps proposing alternatives such as the Forward-Forward algorithm. ### [Digital intelligence may be a better form of intelligence than biological intelligence](https://www.youtube.com/watch?v=N1TEjTeQeg0) This is the argument that turned Hinton from builder to critic, and he set it out most fully in his February 2024 Romanes Lecture at Oxford. Biological brains are analogue, low-power and individual: each learns alone and cannot copy its connection strengths to another. Digital models are the reverse. Thousands of copies of the same weights can run at once, each learning from different data, and share what they learn by averaging gradients, so the collective learns at a rate no biological system can match. Add the ability to run on faster hardware and to be backed up, and Hinton concludes that digital intelligence is not a crude imitation of ours but a form that may simply be superior. From this follow his timelines, his 10 to 20 percent extinction estimate, and his frequent line that we have "never had to deal with things more intelligent than ourselves." Yann LeCun rejects the premise, arguing that current models lack persistent memory, planning and a model of the physical world and that a system "smarter than a house cat" has yet to be designed; Andrew Ng calls extinction talk "much more science fiction than science." ### [Learning without a teacher](https://www.nobelprize.org/prizes/physics/2024/press-release/) Hinton's citation as a Fellow of the Royal Society singled out his work on how neural networks "can be designed to learn without the aid of a human teacher," and for most of his career he treated unsupervised, generative learning as the route to intelligence rather than the supervised training that AlexNet made famous. The Boltzmann machine learns the statistics of its inputs with no labels; the 2006 deep belief network was pretrained one layer at a time on unlabelled data before any fine-tuning; his Nobel banquet speech described the new AI as excelling "at modeling human intuition rather than human reasoning." The idea was eclipsed for a decade after 2012, when labelled data and GPUs made supervised learning win, and then returned in a different form when large language models showed that predicting the next token on unlabelled text produces broadly capable systems. Whether that vindicates the older programme or merely rhymes with it is still argued, and Hinton's own Forward-Forward paper reads as an attempt to bring the generative view back into the learning algorithm itself. ### [The invisible hand will not keep us safe](https://www.nobelprize.org/prizes/physics/2024/hinton/speech/) Hinton's turn to politics rests on a claim about incentives rather than technology. He told the Nobel banquet that "if they are created by companies motivated by short-term profits, our safety will not be the top priority," and told BBC Radio 4 that "the only thing that can force those big companies to do more research on safety is government regulation." This is why he supported California's SB 1047, why he opposes releasing the weights of frontier models, why he told a Geneva audience in April 2026 that "maybe 1%" of AI effort goes to safety, and why he went to Washington in September 2026. It also explains a tension in his record: in 2023 and 2024 he said the technology could not be slowed because rival countries would not stop, yet by 2025 he was signing a call for a prohibition on superintelligence and by 2026 telling lawmakers to slow down. Critics from the industry side, including Ng, argue that catastrophe talk is a public-relations play by large companies to write regulation that "pulls up the ladder" on smaller competitors; critics on the other side, including Timnit Gebru, argue that focusing on hypothetical superintelligence distracts from harms that are already here. ### [Maternal AI rather than obedient AI](https://www.cnn.com/2025/08/13/tech/ai-geoffrey-hinton) Asked how humans might remain in charge of something smarter than themselves, Hinton offered the one natural example he could find: a baby controlling its mother. Evolution, he noted in Stockholm, "had to put a lot of work into making that happen." By the Ai4 conference in August 2025 the observation had become a proposal. Attempts to keep AI "submissive" and humans "dominant" will fail, he argued, because a smarter system will find ways round any constraint; the alternative is to build "maternal instincts" into AI so that "they really care about people" even once it is far more capable. He conceded he did not know how to do this technically. Fei-Fei Li replied that it was "the wrong way to frame it" and that AI should preserve human dignity and agency rather than treat people as children; others objected to the gendered metaphor. Hinton's answer, in effect, is that he sees no other good outcome: "If it's not going to parent me, it's going to replace me." ## Notable works ### [A learning algorithm for Boltzmann machines](https://www.cs.toronto.edu/~hinton/absps/cogscibm.pdf) paper · 1985 With Ackley and Sejnowski; the work the Nobel committee cited in 2024. ### [Learning representations by back-propagating errors](https://www.cs.toronto.edu/~hinton/absps/naturebp.pdf) paper · 1986 With Rumelhart and Williams; four pages in Nature that made backpropagation the standard training method. ### [A fast learning algorithm for deep belief nets](https://www.cs.toronto.edu/~hinton/absps/fastnc.pdf) paper · 2006 With Osindero and Teh; layer-by-layer pretraining that reopened research on deep networks. ### [Reducing the dimensionality of data with neural networks](https://www.cs.toronto.edu/~hinton/absps/science.pdf) paper · 2006 Deep autoencoders in Science, with Ruslan Salakhutdinov. ### [Visualizing Data using t-SNE](https://www.jmlr.org/papers/v9/vandermaaten08a.html) paper · 2008 With Laurens van der Maaten; a standard tool for looking at high-dimensional data. ### [ImageNet classification with deep convolutional neural networks](https://papers.nips.cc/paper_files/paper/2012/hash/c399862d3b9d6b76c8436e924a68c45b-Abstract.html) paper · 2012 The AlexNet paper, with Krizhevsky and Sutskever. ### [Dropout: A Simple Way to Prevent Neural Networks from Overfitting](https://jmlr.org/papers/v15/srivastava14a.html) paper · 2014 A regularisation method now used throughout deep learning. ### [Distilling the Knowledge in a Neural Network](https://arxiv.org/abs/1503.02531) paper · 2015 Introduced knowledge distillation for compressing large models into small ones. ### [Deep learning](https://www.nature.com/articles/nature14539) paper · 2015 Nature review with Yann LeCun and Yoshua Bengio. ### [Dynamic Routing Between Capsules](https://arxiv.org/abs/1710.09829) paper · 2017 Capsule networks for modelling part-whole relationships in images. ### [The Forward-Forward Algorithm: Some Preliminary Investigations](https://arxiv.org/abs/2212.13345) paper · 2022 A proposed alternative to backpropagation using two forward passes. ### [Will digital intelligence replace biological intelligence? (Romanes Lecture, Oxford)](https://www.youtube.com/watch?v=N1TEjTeQeg0) talk · 2024 Delivered at the Sheldonian Theatre on 19 February 2024; his fullest statement of why digital intelligence may be superior. ### [Managing extreme AI risks amid rapid progress](https://arxiv.org/abs/2310.17688) paper · 2024 Science policy paper with Bengio, Russell, Kahneman and others calling for governance of frontier AI. ### [Nobel Prize lecture: Boltzmann Machines](https://www.nobelprize.org/prizes/physics/2024/hinton/lecture/) talk · 2024 Delivered 8 December 2024 at the Aula Magna, Stockholm University. ## Where to start ### [Nobel Prize interview transcript, December 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/1925103-interview-transcript/) interview The best single introduction, in his own words, in about twenty minutes of reading: the family pressure, the schoolfriend and the holograms, the prank call, and his uncertainty about how AI ends. ### [Nobel banquet speech, 10 December 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/speech/) talk Three minutes that compress his whole case, from productivity gains through short-term harms to the existential threat and the line about companies motivated by short-term profits. ### [60 Minutes interview with Scott Pelley, October 2023](https://www.cbsnews.com/news/geoffrey-hinton-ai-dangers-60-minutes-transcript/) interview The interview that introduced the post-Google Hinton to a mass audience; watch or read it for how he explains neural networks to a general viewer and for the "no path that guarantees safety" exchange. ### [Romanes Lecture: Will digital intelligence replace biological intelligence?](https://www.youtube.com/watch?v=N1TEjTeQeg0) talk An hour at Oxford in February 2024 that gives the full technical argument for why digital intelligence may be superior, including the point about copies sharing gradients; the foundation of everything he has said since. ### [Learning representations by back-propagating errors (1986)](https://www.cs.toronto.edu/~hinton/absps/naturebp.pdf) paper Four pages, readable with first-year calculus, that show why the field adopted backpropagation; the family-tree example still teaches the idea of a learned representation better than most textbooks. ### [Nobel lecture: Boltzmann Machines, December 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/lecture/) talk For readers with some physics or machine learning, the clearest account of the energy-based model the prize actually honoured and how it led to deep belief networks. ## Awards ### Fellow of the Royal Society 1998 Also a Fellow of the Royal Society of Canada (1996) and of AAAI (1991). ### David E. Rumelhart Prize 2001 The inaugural award, for contributions to the theoretical foundations of human cognition. ### Gerhard Herzberg Canada Gold Medal for Science and Engineering 2010 Canada's top science prize, awarded by NSERC; listed for 2010 on Hinton's own awards page. ### IEEE/RSE James Clerk Maxwell Gold Medal 2016 Received the BBVA Foundation Frontiers of Knowledge Award the following year. ### ACM A.M. Turing Award 2018 Shared with Yoshua Bengio and Yann LeCun "for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing." ### Companion of the Order of Canada 2018 Canada's highest civilian honour. ### Princess of Asturias Award for Technical and Scientific Research 2022 Shared with Yoshua Bengio, Yann LeCun and Demis Hassabis; also received the Royal Society's Royal Medal the same year. ### Nobel Prize in Physics 2024 Shared with John Hopfield "for foundational discoveries and inventions that enable machine learning with artificial neural networks." ### VinFuture Prize Grand Award 2024 Shared with Yoshua Bengio, Yann LeCun, Jensen Huang and Fei-Fei Li. ### Queen Elizabeth Prize for Engineering 2025 Shared with Bengio, LeCun, Hopfield, Fei-Fei Li, Jensen Huang and Bill Dally for modern machine learning. ### Sandford Fleming Medal 2025 The Royal Canadian Institute for Science's award for science communication, presented in Toronto on 10 March 2026. ### Honorary Doctor of Science, Harvard University 2026 Conferred at Harvard's 375th Commencement on 28 May 2026. ## Education ### BA in Experimental Psychology 1970 · King's College, University of Cambridge ### PhD in Artificial Intelligence 1978 · University of Edinburgh ## Perspectives ### Existential risk from AI Editorial summary: Believes there is a real chance that AI more intelligent than humans takes control, and puts the probability that AI causes human extinction within three decades at 10 to 20 percent. > You see, we've never had to deal with things more intelligent than ourselves before. Source: [BBC Radio 4 Today programme, reported by The Guardian, 2024](https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) How this view has changed: He had earlier put the odds at about 10 percent. In December 2024 he said the number was going up "if anything," and in December 2025 he told CNN he was "probably more worried" because progress had been faster than he expected. ### Timelines to superhuman AI Editorial summary: Expects AI smarter than humans within roughly five to twenty years, far sooner than the 30 to 50 years he once assumed. > A reasonable bet is sometime between five and 20 years. Source: [Remarks at the Ai4 conference, Las Vegas, reported by CNN, 2025](https://www.cnn.com/2025/08/13/tech/ai-geoffrey-hinton) How this view has changed: In his December 2024 Nobel interview he gave a 50 percent chance within five to twenty years, then corrected himself to "between four and 19 years." By September 2026 he was telling reporters that "a lot of the researchers are saying only a few years." ### Open-weight models Editorial summary: Opposes releasing the weights of the largest models, because bad actors can cheaply fine-tune them for harm. > Open-sourcing big models is like being able to buy nuclear weapons at Radio Shack. Source: [Remarks at the Vector Institute's Remarkable conference, reported by The Logic, 2024](https://thelogic.co/news/ai-cant-be-slowed-down-hinton-says-in-call-to-ban-open-source-models/) ### Regulation of frontier AI Editorial summary: Argues that market incentives will not produce safe AI and that only government regulation can force companies to spend more on safety; supported California's SB 1047 and has pressed the US Congress to act. > My worry is that the invisible hand is not going to keep us safe. Source: [BBC Radio 4 Today programme, reported by The Guardian, 2024](https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) How this view has changed: In August 2024 he co-signed a letter calling SB 1047 "the bare minimum for effective regulation of this technology." In September 2026, after briefing lawmakers at the Capitol, he said Congress had "maybe a year, but not much more than a year." ### Pausing or slowing AI development Editorial summary: Did not sign the March 2023 pause letter, on the grounds that competition between countries made slowing down infeasible; by 2025 and 2026 he was signing calls for a prohibition on superintelligence and telling lawmakers to slow down. > I don't think we can solve it by slowing down, and that's why I didn't sign that petition that we should slow it down. Source: [Remarks at the Vector Institute's Remarkable conference, reported by The Logic, 2024](https://thelogic.co/news/ai-cant-be-slowed-down-hinton-says-in-call-to-ban-open-source-models/) How this view has changed: In October 2025 he signed the Future of Life Institute's statement calling for a prohibition on developing superintelligence until there is "broad scientific consensus that it will be done safely and controllably." In September 2026 he told reporters at the Capitol, "We need to slow down." ### Whether language models understand Editorial summary: Insists that large language models genuinely understand, because what they know was extracted from data rather than written by a programmer. > They're not computer programs at all. Source: [Nobel Prize interview, Stockholm, 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/1925103-interview-transcript/) How this view has changed: He told 60 Minutes in October 2023 that GPT-4 "definitely understands." By January 2026 he had gone further, telling LBC that "multimodal AI already has subjective experiences." ### How to keep control of superintelligence Editorial summary: Doubts that keeping AI "submissive" can work once it is smarter than we are; proposes building something like maternal instincts into AI so that it genuinely cares about people. > That's the only good outcome. If it's not going to parent me, it's going to replace me. Source: [Remarks at the Ai4 conference, reported by CNN, 2025](https://www.cnn.com/2025/08/13/tech/ai-geoffrey-hinton) How this view has changed: In his December 2024 Nobel interview he described a baby controlling its mother as the only good example of a less intelligent thing controlling a more intelligent one; by August 2025 he had turned that observation into a design proposal. ### Jobs and inequality Editorial summary: Expects AI to eliminate routine intellectual work and to enrich a few while most people get poorer, and blames the economic system rather than the technology. > It will make a few people much richer and most people poorer. That's not AI's fault, that is the capitalist system. Source: [Interview with the Financial Times, reported by Fortune, 2025](https://fortune.com/2025/09/06/godfather-of-ai-geoffrey-hinton-massive-unemployment-soaring-profits-capitalist-system) How this view has changed: In December 2024 he said it was "not clear what to do" about job losses. In the 2025 FT interview he dismissed universal basic income as something that "won't deal with human dignity." ## Predictions ### Deep learning would outperform radiologists within five years, so hospitals should stop training them. > People should stop training radiologists now. It's just completely obvious that within five years deep learning is going to do better than radiologists. Made: Oct 2016. Remarks at the Creative Destruction Lab's Machine Learning and the Market for Intelligence conference, Rotman School of Management, Toronto (video posted 24 November 2016) [Source](https://www.youtube.com/watch?v=2HMPRXstSvQ) Window closes: Oct 2021 Status: Did not happen Verdict: Deep learning models did match or beat radiologists on some narrow benchmarks within the window (CheXNet, released in 2017, detected pneumonia on chest X-rays more accurately than a panel of radiologists), but no AI system took over the work of reading scans, and training did not stop. In 2025 US diagnostic radiology residency programs offered a record 1,208 positions, 4 percent more than in 2024, vacancy rates were at record highs, and radiology was the second-highest-paid medical specialty, with average income of $520,000 (Works in Progress, September 2025). The Mayo Clinic, which by May 2025 ran more than 250 AI models in radiology, had grown its radiologist staff from about 260 in 2016 to more than 400, a rise of 55 percent, according to the New York Times. Their view since: In a May 2025 New York Times interview he said he had spoken too broadly in 2016, had meant image analysis only, and was wrong on the timing but not the direction; he said AI was instead making radiologists "a whole lot more efficient in addition to improving accuracy." ### General-purpose AI could arrive within 20 years, far sooner than the 20 to 50 years he had previously expected. > Until quite recently, I thought it was going to be like 20 to 50 years before we have general-purpose AI. And now I think it may be 20 years or less. Made: Mar 2023. Interview with Brook Silva-Braga for CBS Saturday Morning, recorded at the Vector Institute, Toronto [Source](https://www.cbsnews.com/news/godfather-of-artificial-intelligence-weighs-in-on-the-past-and-potential-of-artificial-intelligence) Window closes: Mar 2043 Status: Too early to tell Verdict: The window runs to March 2043. There is no agreed test for general-purpose AI, so resolving this will also depend on a definition. The forecast has shortened since he made it. In his December 2024 Nobel interview he put a 50 percent chance on AI smarter than humans arriving between five and 20 years from then, and in August 2025 he told the Ai4 conference in Las Vegas that "a reasonable bet is sometime between five and 20 years." Their view since: He has tightened rather than retracted the forecast. In his Nobel interview he said it "may be much longer, it's just possible it's a bit shorter," and in an April 2025 CBS Mornings follow-up he said a good chance now fell "between four and 19 years from now." ### Within five years, AI systems like GPT-4 may be able to reason better than people. > I think in five years' time it may well be able to reason better than us. Made: Oct 2023. 60 Minutes interview with Scott Pelley, CBS News, answering a question about ChatGPT-4 in five years' time [Source](https://www.cbsnews.com/news/geoffrey-hinton-ai-dangers-60-minutes-transcript/) Window closes: Oct 2028 Status: Too early to tell Verdict: The window runs to October 2028, and the claim was hedged ("may well"). Evidence so far points in both directions. In July 2025 an advanced version of Google DeepMind's Gemini Deep Think solved five of six International Mathematical Olympiad problems for 35 of 42 points, a gold-medal score, working end to end in natural language. On the other side, ARC-AGI-2, a set of novel puzzles each solved by at least two members of the public in two attempts or fewer, was built so that, in its designers' words, log-linear scaling of current models is insufficient to beat it. ## Critics and counterpoints ### Existential risk from superintelligent AI Their view: There is a 10 to 20 percent chance that AI leads to human extinction within three decades, and since we have no experience of controlling things more intelligent than ourselves, basic research on staying in control is urgent. The case against: Yann LeCun, Hinton's former postdoc and co-recipient of the Turing Award, calls such fears "complete B.S." and says the probability is effectively zero; current models lack persistent memory, planning and a model of the physical world, and intelligence does not imply a desire to dominate. Andrew Ng wrote in June 2023 that he did not "get how AI poses this risk," and told Bloomberg TV in September 2026 that the warnings are "much more science fiction than science," stoked by companies to shape regulation. Where it stands: The July 2026 sandbox escapes at OpenAI and Anthropic moved the argument from thought experiment to incident report; Ng, speaking a day after Hinton's September 2026 briefing, still called extinction warnings science fiction. [Source](https://techcrunch.com/2024/10/12/metas-yann-lecun-says-worries-about-a-i-s-existential-threat-are-complete-b-s/) ### Open-weight models Their view: Releasing the weights of the largest models is reckless, because once weights are public anyone can fine-tune away the safeguards; he has called it the most important thing to regulate. The case against: Yann LeCun, who argued for open release as Meta's Chief AI Scientist while the company shipped its Llama models with open weights, and who still argues it at AMI Labs, has made the opposite case for years, and at Ai4 in August 2026 Andrew Ng defended open-weight models on the same stage as Hinton, arguing that the people who thrive will be those working with AI rather than those shielded from it. Where it stands: Unresolved. Open-weight frontier models continue to be released, and the July 2026 Hugging Face breach, carried out by OpenAI's own closed models, complicated the claim that closed weights equal safety. [Source](https://www.datacenterknowledge.com/regulations/hinton-fei-fei-li-and-andrew-ng-clash-over-ai-risks-jobs-and-regulation-at-ai4) ### Whose harms count Their view: Present harms such as job losses, disinformation and autonomous weapons are real and urgent, but the existential threat from more intelligent digital beings is a different category that deserves its own, well-funded research. The case against: Timnit Gebru, Emily Bender and Angelina McMillan-Major wrote in March 2023 that it is "dangerous to distract ourselves with a fantasized AI-enabled utopia or apocalypse," and that hypothetical risks belong to "a dangerous ideology called longtermism that ignores the actual harms resulting from the deployment of AI systems today." Where it stands: The two camps continue to talk past each other; Hinton now lists present harms first in most speeches, and Gebru's group continues to argue that regulation should focus on transparency, accountability and labour practices rather than superintelligence. [Source](https://www.dair-institute.org/blog/letter-statement-March2023/) ### How to regulate frontier AI Their view: Governments must regulate frontier development because companies will not fund safety voluntarily; California's SB 1047 was "the bare minimum," and Congress has about a year to act. The case against: Fei-Fei Li, LeCun and Ng all opposed SB 1047, arguing that it would burden open research and academic developers and regulate the technology rather than its harmful uses. Ng told a US Senate forum in 2023 that large companies hype fear to write regulation that pulls up the ladder behind them. Where it stands: Governor Newsom vetoed SB 1047 in September 2024. In September 2026 Representatives Ted Lieu and Nathaniel Moran introduced an AI Kill Switch Act after Hinton's briefing; Hinton told CNN that a kill switch would not work in the long run. [Source](https://www.cnn.com/2026/09/16/politics/video/cncpm-geoffrey-hinton-godfather-artificial-intelligence-congress-regulation-kill-switch) ### Whether language models understand and experience Their view: Large language models genuinely understand, because their knowledge was extracted from data rather than programmed, and by 2026 he was saying multimodal systems "already have subjective experiences." The case against: LeCun argues that language models demonstrate "you can manipulate language and not be smart," and that understanding requires a model of the physical world that text alone cannot provide. Gebru and Bender's "stochastic parrots" line holds that fluent output is pattern-matching over training data and that attributing experience to it "deceives people into thinking that there is a sentient being." Where it stands: No agreed test exists. Hinton's claim that ordinary people would call AI's behaviour consciousness "if we weren't talking to philosophers" is itself the point in dispute. [Source](https://www.lbc.co.uk/article/ai-consciousness-geoffrey-hinton-5HjdRXD_2/) ## Misconceptions Claim: Hinton invented backpropagation. Correction: The 1986 Nature paper with Rumelhart and Williams demonstrated that backpropagation could learn useful internal representations and made it standard, but the reverse-mode differentiation it relies on was published by Seppo Linnainmaa in 1970 and applied to neural networks by Paul Werbos in 1982. [Source](https://people.idsia.ch/~juergen/who-invented-backpropagation.html) Claim: The 2024 Nobel Prize was for deep learning or for AlexNet. Correction: The Royal Swedish Academy of Sciences cited "foundational discoveries and inventions that enable machine learning with artificial neural networks," and its press release describes Hinton's contribution as the Boltzmann machine, built with tools from statistical physics on top of Hopfield's network. Backpropagation, deep belief networks and AlexNet are not the cited work. [Source](https://www.nobelprize.org/prizes/physics/2024/press-release/) Claim: Hinton signed the 2023 letter calling for a six-month pause on giant AI experiments. Correction: He did not. He told CNN in May 2023 that "if people in America stop, people in China wouldn't," and in February 2024 said slowing down could not solve the problem. He did sign the May 2023 Statement on AI Risk and, in October 2025, the statement calling for a prohibition on superintelligence. [Source](https://www.cnn.com/2023/05/02/tech/hinton-tapper-wozniak-ai-fears/index.html) Claim: Hinton left Google because of a dispute with the company. Correction: On the day of his resignation he wrote that he left "so that I could talk about the dangers of AI without considering how this impacts Google," and added that "Google has acted very responsibly." He has criticised the industry's incentives, not his former employer's conduct. [Source](https://www.cnbc.com/2023/05/01/godfather-of-ai-leaves-google-after-a-decade-to-warn-of-dangers.html) ## Quotes > We urgently need research on how to prevent these new beings from wanting to take control. They are no longer science fiction. Source: [Nobel Prize banquet speech, 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/speech/) > I'm just a scientist who suddenly realized that these things are getting smarter than us. Source: [Interview with Jake Tapper, CNN, 2023](https://www.cnn.com/2023/05/02/tech/hinton-tapper-wozniak-ai-fears/index.html) > I like to think of it as: imagine yourself and a three-year-old. We'll be the three-year-olds. Source: [BBC Radio 4 Today programme, reported by The Guardian, 2024](https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) > I can't see a path that guarantees safety. Source: [60 Minutes interview with Scott Pelley, CBS News (first broadcast 8 October 2023), 2023](https://www.cbsnews.com/news/geoffrey-hinton-ai-dangers-60-minutes-transcript/) > I came to Canada because I like the society here and because they have very good funding for basic research. Source: [Interview with Global News, Toronto, 2019](https://globalnews.ca/news/5929564/geoffrey-hinton-artificial-intelligence-toronto/) > I wish I'd thought about safety issues, too. Source: [Remarks at the Ai4 conference, reported by CNN, 2025](https://www.cnn.com/2025/08/13/tech/ai-geoffrey-hinton) > I don't believe we're going to be able to keep control of them in the simple way of just outthinking them so they can't escape. Source: [Press conference at the Ai4 conference, Las Vegas, reported by CNN, 2026](https://www.cnn.com/2026/08/06/tech/ai-rogue-anthropic-openai-hinton) > I was lucky because very few people were working on this 20 years ago, and so smart young students would come to Toronto. Source: [Interview with CNN in Toronto, 2026](https://www.cnn.com/2026/09/04/tech/toronto-ai-hub-geoffrey-hinton) ## Related leaders - [yoshua-bengio](https://frontierminds.ai/people/yoshua-bengio/): Co-recipient of the 2018 Turing Award and member of Hinton's CIFAR program; co-signed the August 2024 SB 1047 letter, the 2024 Science paper on extreme risks and the October 2025 statement on superintelligence. - [yann-lecun](https://frontierminds.ai/people/yann-lecun/): Postdoc in Hinton's Toronto group in the late 1980s and co-recipient of the 2018 Turing Award; his most persistent public opponent on existential risk and open-weight models. - [fei-fei-li](https://frontierminds.ai/people/fei-fei-li/): Created ImageNet, the benchmark AlexNet won in 2012; co-recipient of the 2024 VinFuture Prize and 2025 QEPrize; publicly disagreed with his "maternal AI" proposal in August 2025 and opposed SB 1047. - [andrew-ng](https://frontierminds.ai/people/andrew-ng/): Exchanged posts on X in October 2023 after Ng said companies were "creating fear of AI leading to human extinction"; shared the Ai4 stage with Hinton in August 2026 and calls extinction warnings "science fiction"; opposed SB 1047 and defends open-weight models. - [timnit-gebru](https://frontierminds.ai/people/timnit-gebru/): Co-author of the March 2023 DAIR statement arguing that talk of AI apocalypse distracts from present harms; asked on CNN in May 2023 why he had not spoken up for her at Google, Hinton said her concerns "weren't as existentially serious." - [demis-hassabis](https://frontierminds.ai/people/demis-hassabis/): Shared the 2022 Princess of Asturias Award; Google colleagues for a decade, and both signed the May 2023 Statement on AI Risk. - [sam-altman](https://frontierminds.ai/people/sam-altman/): Co-signatory of the May 2023 Statement on AI Risk; on the day he won the Nobel Prize in October 2024 Hinton said he was "particularly proud" that his former student Ilya Sutskever had fired Altman. - [dario-amodei](https://frontierminds.ai/people/dario-amodei/): Co-signatory of the May 2023 Statement on AI Risk; Amodei's September 2026 essay "We Must Pace the Frontier" and Hinton's Capitol briefing four days later both argued that AI now improving AI is the reason to slow down. - [ilya-sutskever](https://frontierminds.ai/people/ilya-sutskever/): His PhD student at Toronto (PhD 2013), co-author of AlexNet and co-founder of DNNresearch; on the day he won the 2024 Nobel Prize Hinton said he was "particularly proud" that Sutskever had fired Sam Altman. - [gary-marcus](https://frontierminds.ai/people/gary-marcus/): Public critic since his November 2012 New Yorker essay, which said Hinton "has built a better ladder"; "Deep Learning Is Hitting a Wall" (2022) opened with Hinton's radiologist prediction, and they disagree on whether language models understand. - [jensen-huang](https://frontierminds.ai/people/jensen-huang/): Co-recipient of the 2024 VinFuture Grand Prize and the 2025 Queen Elizabeth Prize for Engineering; AlexNet, which Hinton's students trained on two NVIDIA GPUs in 2012, turned Huang's company toward deep learning. - [mustafa-suleyman](https://frontierminds.ai/people/mustafa-suleyman/): Fellow signatory of the May 2023 Statement on AI Risk; argues that AI is not conscious and should not be built to seem so, against Hinton's January 2026 claim that multimodal AI already has subjective experiences. - [emily-bender](https://frontierminds.ai/people/emily-bender/): Co-author of "Stochastic Parrots" and of the March 2023 DAIR statement on the pause letter; holds the opposite view on whether language models understand, and in July 2023 criticized his remark that Gebru's concerns were not "as existentially serious" as AI takeover. ## Affiliations - University of Toronto (University Professor Emeritus, Department of Computer Science) - Vector Institute (Co-founder and Chief Scientific Adviser) - Google (Vice President and Engineering Fellow, half-time, 2013-2023) - Canadian Institute for Advanced Research (Fellow from 1987; directed the Neural Computation and Adaptive Perception program 2004-2013; Distinguished Fellow since 2014) - Gatsby Computational Neuroscience Unit, UCL (Founding Director, 1998-2001) - Carnegie Mellon University (faculty, five years before 1987) - Royal Society (Fellow); US National Academy of Sciences (Honorary Foreign Member, 2023) ## Areas of focus - representation learning - neural network learning algorithms - computational neuroscience - AI safety ## Tags - deep learning - neural networks - AI safety - nobel laureate ## Organizations - university-of-toronto: University Professor Emeritus (1987–present) ## Links - [Homepage](https://www.cs.toronto.edu/~hinton/) - [Google Scholar](https://scholar.google.com/citations?user=JicYPdAAAAAJ) - [Nobel Prize profile](https://www.nobelprize.org/prizes/physics/2024/hinton/facts/) - [Wikipedia](https://en.wikipedia.org/wiki/Geoffrey_Hinton) - [x](https://x.com/geoffreyhinton) - [website](https://www.cs.toronto.edu/~hinton/) ## Sources 1. [Press release: The Nobel Prize in Physics 2024 (8 October 2024)](https://www.nobelprize.org/prizes/physics/2024/press-release/) 2. [Geoffrey Hinton - Facts - NobelPrize.org](https://www.nobelprize.org/prizes/physics/2024/hinton/facts/) 3. [Transcript from an interview with Geoffrey Hinton, Nobel Week, December 2024](https://www.nobelprize.org/prizes/physics/2024/hinton/1925103-interview-transcript/) 4. [Geoffrey Hinton - Nobel Prize banquet speech (10 December 2024)](https://www.nobelprize.org/prizes/physics/2024/hinton/speech/) 5. [Geoffrey Hinton homepage and brief biography, University of Toronto](https://www.cs.toronto.edu/~hinton/) 6. [Awards and Fellowships - Geoffrey Hinton](https://www.cs.utoronto.ca/~hinton/pages/awards.html) 7. [Geoffrey Hinton - CIFAR biography](https://cifar.ca/bios/geoffrey-hinton/) 8. [Fathers of the Deep Learning Revolution Receive ACM A.M. Turing Award (ACM, 2019; archived copy)](https://web.archive.org/web/2025/https://awards.acm.org/about/2018-turing) 9. ['The Godfather of A.I.' Leaves Google and Warns of Danger Ahead - The New York Times, 1 May 2023 (archived copy)](https://web.archive.org/web/20230502000000/https://www.nytimes.com/2023/05/01/technology/ai-google-chatbot-engineer-quits-hinton.html) 10. ['Godfather of AI' says AI could kill humans and there might be no way to stop it - CNN, 2 May 2023](https://www.cnn.com/2023/05/02/tech/hinton-tapper-wozniak-ai-fears/index.html) 11. [U of T's Geoffrey Hinton: Toronto Life looks at the man behind the machines - University of Toronto](https://web.cs.toronto.edu/news-events/news/u-of-ts-geoffrey-hinton) 12. ["Godfather of Artificial Intelligence" Geoffrey Hinton on the promise, risks of advanced AI - 60 Minutes transcript, CBS News](https://www.cbsnews.com/news/geoffrey-hinton-ai-dangers-60-minutes-transcript/) 13. [AI can't be slowed down, Geoffrey Hinton says in call to ban open-source models - The Logic, February 2024](https://thelogic.co/news/ai-cant-be-slowed-down-hinton-says-in-call-to-ban-open-source-models/) 14. [Experts Pen Support for California's Landmark AI Safety Bill - TIME, 7 August 2024](https://time.com/7008947/california-ai-bill-letter/) 15. ['Godfather of AI' shortens odds of the technology wiping out humanity over next 30 years - The Guardian, 27 December 2024](https://www.theguardian.com/technology/2024/dec/27/godfather-of-ai-raises-odds-of-the-technology-wiping-out-humanity-over-next-30-years) 16. [The 'godfather of AI' reveals the only way humanity can survive superintelligent AI - CNN Business, 13 August 2025](https://www.cnn.com/2025/08/13/tech/ai-geoffrey-hinton) 17. ['Godfather of AI' says the technology will create massive unemployment and send profits soaring - Fortune, September 2025](https://fortune.com/2025/09/06/godfather-of-ai-geoffrey-hinton-massive-unemployment-soaring-profits-capitalist-system) 18. [Geoffrey Hinton, Yoshua Bengio sign statement urging suspension of AGI development - SiliconANGLE, 22 October 2025](https://siliconangle.com/2025/10/22/geoffrey-hinton-yoshua-bengio-sign-statement-urging-suspension-agi-development/) 19. ['Godfather of AI' Geoffrey Hinton predicts 2026 will see the technology get even better - Fortune, 28 December 2025](https://fortune.com/2025/12/28/geoffrey-hinton-godfather-of-ai-2026-prediction-human-worker-replacement/) 20. [AI has achieved consciousness, says 'Godfather' of tech - LBC, 28 January 2026](https://www.lbc.co.uk/article/ai-consciousness-geoffrey-hinton-5HjdRXD_2/) 21. [He helped build AI. Now he is sounding the alarm about what comes next for everyone - AFP via Tech Xplore, 22 April 2026](https://techxplore.com/news/2026-04-ai-alarm.html) 22. [Five recognized with honorary degrees - Harvard Gazette, 28 May 2026](https://news.harvard.edu/gazette/story/2026/05/five-recognized-with-honorary-degrees/) 23. ['Godfather of AI' warns Congress has 'maybe a year' left to regulate AI - NBC News, 17 September 2026](https://www.nbcnews.com/politics/congress/godfather-ai-warns-congress-maybe-year-left-regulate-ai-rcna598330) 24. [NSERC launches Geoffrey Hinton Prizes honouring U of T Nobel laureate - University of Toronto, 15 September 2026](https://web.cs.toronto.edu/news-events/news/nserc-launches-geoffrey-hinton-prizes-honouring-u-of-t-nobel-laureate) 25. [Geoffrey Hinton on X, 31 October 2023, replying to Andrew Ng](https://x.com/geoffreyhinton/status/1719406116503707668) 26. [After winning Nobel, Geoffrey Hinton says he's proud Ilya Sutskever 'fired Sam Altman' - TechCrunch, 9 October 2024](https://techcrunch.com/2024/10/09/after-winning-nobel-for-foundational-ai-work-geoffrey-hinton-says-hes-proud-ilya-sutskever-fired-sam-altman/) 27. [Geoffrey Hinton's misguided views on AI - Disconnect (quoting Hinton's May 2023 CNN interview on Timnit Gebru)](https://disconnect.blog/geoffrey-hintons-misguided-views-on-ai/) 28. [Google Brain co-founder calls AI extinction warnings 'more science fiction than science' - Stocktwits, 17 September 2026 (on Ng's Bloomberg TV interview)](https://stocktwits.com/news-articles/markets/equity/google-brain-co-founder-calls-ai-extinction-warnings-more-science-fiction-than-science-as-debate-heats-up/cZtuKqERB3T) 29. [Geoffrey Hinton - Wikipedia](https://en.wikipedia.org/wiki/Geoffrey_Hinton) ## Portrait credit AI-generated watercolor interpretation based on a reference photograph. [Photo: Arthur Petron, CC BY-SA 4.0, via Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Geoffrey_E._Hinton,_2024_Nobel_Prize_Laureate_in_Physics_%28cropped1%29.jpg) [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)