Geoffrey Hinton

Geoffrey Everest Hinton was born on December 6, 1947, in Wimbledon, on the southwestern edge of London. Science ran deep in the family. His father, Howard Everest Hinton, was an entomologist and fellow of the Royal Society, and his great-great-grandfather was George Boole, the mathematician whose logic underpins digital computing. His middle name honours a relative, the surveyor George Everest, after whom the mountain is named ^[1].

Hinton grew up in Bristol, where his father taught at the university, and attended Clifton College. The household set high expectations: he later recalled being told from an early age that he was expected to become an academic. At King's College, Cambridge, he wandered between subjects, trying physiology, physics and philosophy before finishing a degree in experimental psychology in 1970 ^[2].

The question that gripped him as a student was how the brain stores and processes information. Neither the physiologists nor the philosophers at Cambridge could answer it to his satisfaction, and for a period after graduating he worked as a carpenter while deciding what to do next. The unresolved puzzle eventually pulled him back into research ^[3].

In 1972 Hinton began a PhD in artificial intelligence at the University of Edinburgh under Christopher Longuet-Higgins. His timing looked terrible. The field had turned against neural networks, brain-inspired systems of simple connected units, after an influential 1969 critique by Marvin Minsky and Seymour Papert, and funding for the approach had largely dried up. Hinton persisted anyway, convinced that learning in networks of neuron-like units was the right model for intelligence, and completed his doctorate in 1978 ^[2].

Unable to find support for this line of work in Britain, he moved to the United States, first as a postdoctoral researcher at the University of California, San Diego, where he joined a group of cognitive scientists studying parallel distributed processing, and then to Carnegie Mellon University in Pittsburgh in 1982 ^[1].

The decisive result came in 1986. With David Rumelhart and Ronald Williams, Hinton published a paper in Nature showing that the backpropagation algorithm could train multi-layer neural networks to discover useful internal representations of data ^[4]. Backpropagation works by measuring a network's error and passing corrections backwards through its layers, adjusting the strength of each connection. The paper gave researchers a practical way to train networks deeper than a single layer, removing the obstacle Minsky and Papert had identified. During the same period Hinton and Terrence Sejnowski invented the Boltzmann machine, an early learning network grounded in statistical physics ^[5].

In 1987, uneasy about US military funding of AI research, Hinton moved to the University of Toronto and became a fellow of the Canadian Institute for Advanced Research. Canada became his permanent base, and Toronto's computer science department became the unlikely capital of neural network research through the long years when few others believed in it ^[3].

Any account of Geoffrey Hinton achievements has to reckon with how unfashionable his field was for most of his career. Through the 1990s and 2000s, mainstream machine learning favoured other statistical methods, and neural network papers struggled to get accepted at major conferences. Hinton kept producing ideas anyway: mixtures of experts, Helmholtz machines, variational methods for learning, and in 2006 a fast training scheme for deep belief networks that helped revive interest in many-layered models, work that gave the field its modern name, deep learning ^[5].

The turning point arrived in 2012. Hinton and two of his Toronto graduate students, Alex Krizhevsky and Ilya Sutskever, entered the ImageNet Large Scale Visual Recognition Challenge with a deep convolutional network trained on graphics processors. Their system, later known as AlexNet, cut the error rate so dramatically that the entire computer vision community changed course within months ^[6]. The trio formed a small company, DNNresearch, which Google acquired in 2013. Hinton then split his time between the University of Toronto and Google, where he worked at Google Brain as a vice president and engineering fellow ^[3].

Recognition followed on a grand scale. In 2018 Hinton shared the ACM A.M. Turing Award, computing's highest honour, with Yoshua Bengio and Yann LeCun for conceptual and engineering breakthroughs that made deep neural networks a core component of computing; the three are widely described as the godfathers of deep learning ^[7]. In October 2024 the Royal Swedish Academy of Sciences awarded him the Nobel Prize in Physics, shared with John Hopfield, for foundational discoveries that enable machine learning with artificial neural networks, citing his work on the Boltzmann machine ^[5]. He is also a fellow of the Royal Society, a Companion of the Order of Canada, and a recipient of Canada's Killam Prize, among many other honours ^[1].

Hinton has spoken openly about the losses that shaped his adult life. His first wife, Rosalind Zalin, died of ovarian cancer in 1994, leaving him to raise their two adopted children. His second wife, the art historian Jackie Ford, died of pancreatic cancer in 2018. He has said that the experience of caring for family through serious illness influenced his interest in applying machine learning to medicine ^[3].

A back condition has kept him from sitting down comfortably since the mid-2000s, so for years he worked standing up and avoided flying whenever possible, travelling between Toronto and other cities by train or lying down in vehicles. Colleagues describe a dry, self-deprecating wit and a habit of explaining hard ideas through vivid analogies ^[3].

He holds both British and Canadian citizenship and has lived in Toronto since the late 1980s. Several relatives besides Boole were notable in their own right, including the surgeon James Hinton and the mathematician Charles Howard Hinton, who wrote about the fourth dimension, a lineage Hinton mentions with amusement rather than reverence ^[1].

In May 2023 Hinton resigned from Google at age 75, explaining that he wanted to be able to discuss the dangers of artificial intelligence without considering how it affected his employer. In interviews he said that part of him regretted his life's work, and that the arrival of large language models had shortened his estimate of when machines might exceed human intelligence ^[8].

Since leaving Google he has become one of the most quoted voices on AI risk. His stated concerns range from near-term harms, such as floods of synthetic misinformation, job displacement and autonomous weapons, to the longer-term possibility that systems more capable than people could escape human control. He has supported calls for governments to fund safety research and has argued that companies developing the most powerful models should devote a substantial share of their computing resources to safety work ^[8].

He remains University Professor Emeritus at the University of Toronto and continues to advise the Vector Institute for Artificial Intelligence, which he co-founded in Toronto in 2017. The Nobel announcement in 2024 reached him in a California hotel room; at the news conference that followed he used the platform, characteristically, to repeat his warnings about the technology being honoured ^[5].

Who was Geoffrey Hinton to the field he transformed? For readers approaching a Geoffrey Hinton biography, the simplest answer is that he kept a scientific idea alive through decades of rejection and then watched it remake the world. Backpropagation, distributed representations, dropout regularization and deep belief networks all trace to his group, and the students and postdocs he trained, among them Yann LeCun, Ilya Sutskever, Alex Krizhevsky, Ruslan Salakhutdinov and Yee Whye Teh, populate the senior ranks of AI research and industry ^[7].

The technologies his work enabled now sit inside speech recognition, machine translation, medical imaging, and the chatbots used by hundreds of millions of people. Toronto's rise as an AI hub, anchored by the Vector Institute, is a direct product of his decision to settle there in 1987 ^[6].

His legacy also includes a rarer act: a scientist at the summit of his field publicly questioning where it is heading. Whether history records him mainly as the architect of deep learning or as the insider who sounded the alarm, the basic Geoffrey Hinton facts are settled. He bet on the brain as a guide to building intelligent machines when almost nobody else would, and he won the bet more completely than even he expected ^[8].