Pioneers of AI
The people behind the ideas, from Lovelace to Hassabis, including the many whose work was ignored, under-credited or rediscovered decades later. Plus the Turing Award roll and the field’s fiercest credit dispute.
In 1970 a Finnish master’s student, Seppo Linnainmaa, described an efficient way to compute how every step of a long calculation contributes to its final result. He was not thinking about neural networks. Sixteen years later, a version of the same idea, under the name backpropagation, became the method that trains essentially every neural network in use today. Ask most people who invented it and they will name someone else. That is the problem with the history of any science: the story remembers the people who made an idea famous, and often forgets those who had it first. This page tries to remember both.
Why names matter, and mislead
Histories of AI are usually told through a few founding fathers: Turing, McCarthy, Minsky, Simon. They earned their place. But ideas rarely have single inventors. They emerge in several places at once, are published in the wrong journal or the wrong language, or wait years for the hardware that makes them useful. Women, researchers outside the United States and Britain, and people working in neighbouring fields such as linguistics or statistics are especially easy to overlook.
This guide’s framing The cards below are grouped by the era in which each person did their best-known work. Each gives one or two sentences on what they contributed and a link to the history page where you will meet them in context. Tap a name to open it.
The pioneers, era by era
1837Charles BabbageMathematician and engineer, England
Designed the Difference Engine and then the Analytical Engine, the first design for a general-purpose programmable computer, though neither was completed in his lifetime. Babbage and Lovelace →
1843Ada LovelaceMathematician, England
Her 1843 Notes on the Analytical Engine include what is often called the first published computer program and the insight that such a machine could process symbols of any kind. She also gave the first sceptical argument about machine creativity: the Engine “has no pretensions whatever to originate anything”. Babbage and Lovelace →
ATAlan TuringMathematician, England
Defined computation itself with the Turing machine (1936), helped break German ciphers at Bletchley Park, and in 1950 proposed the imitation game as a way to approach the question “Can machines think?” Alan Turing →
M&PWarren McCulloch & Walter PittsNeurophysiologist and logician, USA
Their 1943 paper modelled neurons as simple on–off logic units and showed networks of them could compute, the first theory of a neural network. Neurons, feedback and information →
NWNorbert WienerMathematician, USA
Founded cybernetics (1948), the study of feedback and control in animals and machines, and was among the first to warn publicly about automation’s effect on work. Neurons, feedback and information →
CSClaude ShannonMathematician and engineer, USA
Created information theory (1948), wrote a founding paper on computer chess (1950), built a maze-learning mechanical mouse, and co-wrote the Dartmouth proposal. Neurons, feedback and information →
JvNJohn von NeumannMathematician, Hungary and USA
His 1945 EDVAC report described the stored-program architecture most computers still use; his last, unfinished lectures were published as The Computer and the Brain (1958). Neurons, feedback and information →
JMJohn McCarthyComputer scientist, USA
Coined the term “artificial intelligence” in the 1955 Dartmouth proposal, invented LISP (1958) and founded the Stanford AI Laboratory. Dartmouth 1956 →
MMMarvin MinskyCognitive scientist, USA
Co-founded MIT’s AI laboratory, proposed “frames” for representing knowledge, and with Seymour Papert wrote Perceptrons (1969), which exposed the limits of single-layer networks. The Golden Years →
N&SAllen Newell & Herbert SimonCarnegie Tech (now Carnegie Mellon), USA
With Cliff Shaw they wrote the Logic Theorist (1956) and the General Problem Solver, and argued that intelligence is the manipulation of symbols. Simon also won the 1978 Nobel Prize in economics for work on decision-making. The Golden Years →
FRFrank RosenblattPsychologist, USA
Built the perceptron (1958), a machine that learned to classify patterns from examples. He died in 1971, before neural networks were vindicated. The Golden Years →
ASArthur SamuelComputer scientist, IBM, USA
His checkers program improved by playing against itself, and his 1959 paper popularised the phrase “machine learning”. The Golden Years →
JWJoseph WeizenbaumComputer scientist, Germany and USA
Wrote ELIZA (1966), then became one of AI’s sharpest critics after seeing how readily people confided in it, in Computer Power and Human Reason (1976). Lessons from History →
EFEdward FeigenbaumComputer scientist, USA
Led Dendral and championed the expert system, arguing that a program’s power comes mainly from the specialist knowledge it holds. Expert systems →
JPJudea PearlComputer scientist, Israel and USA
Developed Bayesian networks for reasoning under uncertainty (1988) and later a mathematics of cause and effect. The statistical turn →
GHGeoffrey HintonCognitive psychologist and computer scientist, UK and Canada
Co-author of the 1986 backpropagation paper, co-inventor of the Boltzmann machine, and supervisor of the AlexNet team in 2012. Shared the 2024 Nobel Prize in Physics with John Hopfield. The connectionist revival →
YLYann LeCunComputer scientist, France and USA
Developed convolutional networks trained with backpropagation at Bell Labs (1989), used to read handwritten cheques and postcodes. The connectionist revival →
YBYoshua BengioComputer scientist, Canada
His 2003 neural language model learned word embeddings and next-word prediction together; his lab’s 2014 work on attention fed directly into the Transformer. Attention Is All You Need →
S&BRichard Sutton & Andrew BartoComputer scientists, USA and Canada
Laid the foundations of modern reinforcement learning, including temporal-difference learning, and wrote its standard textbook (1998). Lessons from History →
RBRodney BrooksRoboticist, Australia and USA
Argued that robots should be built from layers of simple behaviours reacting to the world rather than from elaborate internal models, and co-founded iRobot, maker of the Roomba. The connectionist revival →
VVVladimir VapnikMathematician, USSR and USA
Co-developed statistical learning theory in the Soviet Union with Alexey Chervonenkis, and with Corinna Cortes the support-vector machine (1995). The statistical turn →
DHDemis HassabisNeuroscientist and entrepreneur, UK
Co-founded DeepMind in 2010 with Shane Legg and Mustafa Suleyman; led the teams behind AlphaGo and AlphaFold, and shared the 2024 Nobel Prize in Chemistry. The deep learning revolution →
JJJohn JumperChemist and computer scientist, USA
Led the AlphaFold 2 team at DeepMind and shared the 2024 Nobel Prize in Chemistry with Hassabis (and David Baker). Attention Is All You Need →
ISIlya SutskeverComputer scientist, Russia, Israel and Canada
Co-author of AlexNet (2012) and sequence-to-sequence learning (2014), and a co-founder and long-time chief scientist of OpenAI. The deep learning revolution →
IGIan GoodfellowComputer scientist, USA
Proposed generative adversarial networks (2014), in which two networks compete to produce ever more realistic images. The deep learning revolution →
AV+7Ashish Vaswani and co-authorsGoogle, 2017
Vaswani, Shazeer, Parmar, Uszkoreit, Jones, Gomez, Kaiser and Polosukhin wrote “Attention Is All You Need”, listed in random order as equal contributors. Most later left Google to found or join AI companies. Attention Is All You Need →
The overlooked
These people appear less often in popular histories than their work deserves. Some were women in a field that rarely credited them; some published in Russian, Japanese or Finnish, or in fields outside computer science; some were simply early.
MMMargaret MastermanLinguist and philosopher, Cambridge
Founded the Cambridge Language Research Unit in 1955, one of the first groups to work on machine translation and computational meaning, and an early proponent of semantic networks. The Golden Years →
KSJKaren Spärck JonesComputer scientist, Cambridge
Trained at Masterman’s unit, she introduced inverse document frequency in 1972: weighting words by how rare they are across documents. It became a core ingredient of search engines. The statistical turn →
AIAlexey IvakhnenkoMathematician, Soviet Ukraine
With Valentin Lapa (1965) he trained multilayer networks layer by layer from data, which some historians count as the first deep learning. The Complete AI Timeline →
SAShun’ichi AmariMathematician, Japan
In 1967 he described training a multilayer network by stochastic gradient descent, and in 1972 a self-organising associative memory related to the later Hopfield network. The connectionist revival →
SLSeppo LinnainmaaMathematician, Finland
His 1970 master’s thesis set out reverse-mode automatic differentiation, the general algorithm of which backpropagation is a special case. The connectionist revival →
KFKunihiko FukushimaEngineer, Japan
His neocognitron (1979–80), inspired by the visual cortex, stacked layers of feature detectors in the arrangement later used by convolutional networks. The connectionist revival →
SHSepp HochreiterComputer scientist, Germany and Austria
His 1991 diploma thesis analysed the “vanishing gradient” that stopped deep and recurrent networks learning; with Jürgen Schmidhuber he then designed the LSTM (1997). Attention Is All You Need →
FFLFei-Fei LiComputer scientist, China and USA
Led the creation of ImageNet when many colleagues doubted that a giant labelled dataset was worth building; its 2012 challenge launched the deep-learning era. The deep learning revolution →
GSJoy Buolamwini & Timnit GebruComputer scientists, USA
Their 2018 study Gender Shades found commercial face-analysis systems misclassified darker-skinned women up to 34.7% of the time, against at most 0.8% for lighter-skinned men. It made algorithmic bias impossible to ignore. The Generative Era → · What AI Gets Wrong →
CBCynthia BreazealRoboticist, USA
Built Kismet at MIT in the late 1990s, a robot head that read and expressed social cues, and helped found the field of social robotics. The statistical turn →
DKDaphne KollerComputer scientist, Israel and USA
A leader in probabilistic graphical models (her 2009 textbook with Nir Friedman is a standard), she co-founded Coursera in 2012 and later applied machine learning to drug discovery. The statistical turn →
Who gets the credit?
No credit dispute in AI is louder than the one pressed by Jürgen Schmidhuber, long associated with the Swiss AI lab IDSIA and co-inventor of the LSTM. For years he has argued, in long surveys and public comments, that the deep-learning story as usually told under-credits earlier work: Linnainmaa’s 1970 algorithm, Ivakhnenko’s 1965 deep networks, Amari’s 1967 gradient training, Fukushima’s neocognitron, and work from his own lab, including 1990s ideas he describes as precursors of generative adversarial networks and of the “linear” variants of the Transformer.
Debated Some of his specific citations are widely accepted as historically accurate; whether they mean the famous names deserve less credit is a matter of sharp disagreement.
- Priority matters: Schmidhuber and supporters argue that science should credit whoever first published an idea, and that prizes and press have rewritten history around a few well-connected labs.
- Making it work matters: others reply that an idea counts most when someone shows it works and persuades the field, as Rumelhart, Hinton and Williams did for backpropagation in 1986, and that independent rediscovery is common and honest.
- Many hands: a third view, close to this guide’s, holds that the “inventor” question is often the wrong one: breakthroughs come from chains of contributors, plus data and hardware nobody individually invented.
AI and the Turing Award
The ACM A.M. Turing Award, often called the Nobel Prize of computing, has gone to AI research several times. The year is the award year; since the 2000s the announcement usually comes the following spring.
| Award year | Recipients | Cited for (in brief) |
|---|---|---|
| 1969 | Marvin Minsky | Central role in creating and shaping AI |
| 1971 | John McCarthy | Major contributions to AI |
| 1975 | Allen Newell, Herbert Simon | AI, the psychology of human cognition and list processing |
| 1994 | Edward Feigenbaum, Raj Reddy | Large-scale AI systems, including expert systems and speech recognition |
| 2010 | Leslie Valiant | Theory of computation, including the theory of learning (PAC learning) |
| 2011 | Judea Pearl | Probabilistic and causal reasoning |
| 2018 | Yoshua Bengio, Geoffrey Hinton, Yann LeCun | Making deep neural networks a critical part of computing |
| 2024 | Andrew Barto, Richard Sutton | Foundations of reinforcement learning |
AI-related awards only, up to the 2024 award; the 2025 award went to Charles Bennett and Gilles Brassard for quantum cryptography As of 2026. Herbert Simon (economics, 1978) and Geoffrey Hinton (physics, 2024) are also Nobel laureates; Demis Hassabis and John Jumper shared the 2024 Chemistry prize.
Who introduced inverse document frequency, a building block of search engines?
Spärck Jones published it in 1972. Masterman founded the unit where she trained.
Whose 1979–80 neocognitron anticipated convolutional networks?
Fukushima’s layered vision model is a direct ancestor of LeCun’s convolutional networks.
Which pair shared the 2024 ACM Turing Award for reinforcement learning?
Hassabis and Jumper (Chemistry) and Hopfield and Hinton (Physics) won Nobel Prizes in 2024.
// Pause and reflect
- Should credit go to the person who first published an idea, or the one who made it work and convinced others? Does your answer change for different cases on this page?
- Why might work published in Russian, Japanese or Finnish, or by women in the 1950s–70s, have been overlooked? Are similar blind spots operating today?
- Which pioneer would you most like to ask about today’s AI, and what would you ask?
Your notes are saved privately in this browser.
What happened next
Pioneers made predictions as well as discoveries, and many of them were wrong in revealing ways. The final page of the story, Lessons from History, asks what their hits and misses teach us about today’s claims.
Key terms on this page
Sources & further reading
- Lovelace, A. A., Notes to Sketch of the Analytical Engine invented by Charles Babbage (1843).
- Spärck Jones, K., A statistical interpretation of term specificity and its application in retrieval, Journal of Documentation 28 (1972).
- Buolamwini, J. & Gebru, T., Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification (2018).
- Schmidhuber, J., Annotated History of Modern AI and Deep Learning (2022).
- ACM, A.M. Turing Award laureates by year.
- McCorduck, P., Machines Who Think (1979; 2nd edn 2004).
- Metz, C., Genius Makers (2021).