AI History Project

AI Data Visualizations

aihistoryproject.org

Paradigm Shifts

Eight decades of AI research, grouped into the paradigms that defined them — symbolic AI rising and collapsing, the statistical turn, and the deep-learning era that swallowed the field.

VIEW:

Each decade normalised to 100% — paradigm mix, free of the uneven item counts per decade.

Foundations & Cyberneticspeak 1940s
The theoretical groundwork — computability, feedback, and the mind as a machine.
Symbolic AI & Knowledgepeak 1960s
Intelligence as explicit rules and symbols — logic, expert systems, knowledge bases.
Statistical ML & Datapeak 2000s
The statistical turn — learning patterns from data rather than hand-coded rules.
Neural Nets & Deep Learningpeak 2010s
Connectionism and its revivals — perceptrons, backpropagation, deep architectures.
Generative AI & LLMspeak 2020s
Scale, pretraining and generation — language models, agents, and reasoning.
Perception, Language & Roboticspeak 2010s
The applied frontier every paradigm attacked — seeing, speaking, and moving.
Compute & Infrastructurepeak 1940s
The substrate — the machines, chips, and systems each era could run on.
Safety, Ethics & Governancepeak 2010s
The consequences thread — alignment, bias, policy, and who answers for it.
What the Shape Shows
30% → ~0
Symbolic AI’s arc
Takes nearly a third of the 1970s, then all but vanishes. One 2010s entry inherits it — knowledge graphs — and it survives as infrastructure rather than as a research programme.
1990s
The statistical turn
Statistical ML & Data reaches its high point as learning from data displaces hand-coded rules.
Twice
Neural nets return
A 1980s backpropagation revival, a dip through the 2000s, then the 2010s deep-learning surge.
40%
The generative era
Generative AI & LLMs go from nothing before the 1980s to the largest share of the 2020s.
Derived from the tags on all 446 timeline items · 426 carry at least one paradigm tagHover a band or a card to isolate it
aihistoryproject.org

The Model Size Explosion

From 1 parameter to 1.76 trillion — how AI models grew 1,760,000,000,000× in 65 years. A logarithmic journey through the exponential growth of artificial intelligence.

Neural Net
CNN
Transformer
LLM
Built for aihistoryproject.org — To understand where AI is going, start with where it began.Sources: Epoch AI · Stanford AI Index · Original papers
Loading Map...
aihistoryproject.org

AI Careers in Motion

How 25 personalities moved across universities, labs, and companies — revealing the hidden network behind AI's biggest breakthroughs.

Sort:
Filter:
1980s1990s2000s2010s2020sNowGeoffrey HintonCarnegie …University of TorontoGoogle BrainIndependentJensen HuangOregon St…LSI Logic / AMDNVIDIAYann LeCunSorbonne …AT&T Bell LabsNew York UniversityMeta AI (FAIR)Yoshua BengioMcGill Un…University of Montreal…Jeff DeanCarnegie …DEC Resea…GoogleAndrew NgCarnegie …UC BerkeleyStanford UniversityBaiduLanding AI / DeepLearn…Demis HassabisUniversit…Lionhead / Elixir StudiosUniversit…DeepMindFei-Fei LiPrinceton…CaltechStanford UniversityDavid SilverUniversit…Elixir St…Universit…DeepMindShane LeggIDSIA (Sw…Universit…DeepMindDario AmodeiStanford …Princeton…D.E. Shaw…OpenAIAnthropicSam AltmanLooptY CombinatorOpenAIMustafa SuleymanDeepMindGoogleIlya SutskeverUniversity of TorontoGoogle BrainOpenAIAndrej KarpathyUniversit…Stanford …TeslaEureka LabsAlex KrizhevskyUniversity of TorontoGoogleIndependentDaniela AmodeiStripeOpenAIAnthropicMira MuratiDartmouth…Goldman S…OpenAIIan GoodfellowUniversit…AppleDeepMindAshish VaswaniColumbia …Google BrainEssential AIJohn SchulmanUC BerkeleyOpenAIClement DelangueVarious s…Hugging FaceAlec RadfordOpenAIArthur MenschENS / InriaGoogle De…Mistral AIAidan GomezUniversit…Cohere
Organization Groups
Academic(23)
Google / DeepMind(14)
OpenAI(9)
Anthropic(3)
Meta(1)
Microsoft(1)
NVIDIA(1)
Tesla(2)
Other Industry(9)
Startups / Independent(15)
Key Patterns
The Hinton Tree
1 professor → 3 revolutions
Sutskever (OpenAI/SSI), Krizhevsky (AlexNet), LeCun (Meta) all trained under Hinton at Toronto.
The OpenAI Diaspora
6 senior exits
Sutskever → SSI, Amodeis → Anthropic, Schulman → Anthropic, Karpathy → Eureka, Murati → startup.
Google as Finishing School
14 of 25 passed through
More than half of AI's top figures worked at Google/Brain/DeepMind at some point.
University of Toronto
The #1 AI feeder
Hinton, LeCun, Sutskever, Karpathy, Krizhevsky, Gomez — one department, six world-changers.
Built for aihistoryproject.org — To understand where AI is going, start with where it began.Click any name to explore · Hover to highlight connections
aihistoryproject.org

Academic Lineage

Who taught whom. Doctoral advisors traced from Wikidata, revealing the schools AI grew out of — and that the field has no single family tree, but 14 separate ones.

On the timeline (85) — click to openTeachers outside it (98)· advisor → student, left to right
The John McCarthy line8 on the timeline · 17 people
John McCarthyDonald C. SpencerJohn Edensor LittlewoodG.H. HardyFrank SmithiesSeymour PapertJean PiagetOtto FuhrmannGerald Jay SussmanDrew McDermottTerry WinogradBarbara LiskovSanjay GhemawatFrans KaashoekRaj ReddyJames K. BakerHans Moravec
The David Rumelhart line7 on the timeline · 12 people
Andrew NgMichael I. JordanDonald NormanR. Duncan LucePercy LiangDan KleinDavid RumelhartWilliam Kaye EstesB. F. SkinnerQuoc V. LePieter AbbeelJohn Schulman
The Herbert A. Simon line5 on the timeline · 9 people
Allen NewellHerbert A. SimonHarold LasswellCharles Edward MerriamHenry SchultzHenry Ludwell MooreEdward FeigenbaumDouglas LenatRichard Fikes
The Geoffrey Hinton line5 on the timeline · 8 people
Geoffrey HintonH. Christopher Longuet-HigginsCharles CoulsonYee Whye TehAidan GomezYarin GalRuslan SalakhutdinovIlya Sutskever
The John Holland line4 on the timeline · 8 people
Andrew BartoBernard P. ZeiglerJohn HollandArthur BurksCooper Harold LangfordRichard SuttonDavid SilverMartin Müller
The Fei-Fei Li line3 on the timeline · 7 people
Fei-Fei LiPietro PeronaJitendra MalikChristof KochValentino BraitenbergTimnit GebruAndrej Karpathy
The Leslie Kaelbling line2 on the timeline · 7 people
Stanley RosenscheinAravind JoshiDavid ChiangAshish VaswaniLiang HuangLeslie KaelblingNils John Nilsson
The Alain Colmerauer line2 on the timeline · 6 people
Philippe JorrandLouis BollietJean KuntzmannAlain ColmerauerGeorges ValironRémi Coulom
The Jürgen Schmidhuber line2 on the timeline · 5 people
Klaus SchultenJürgen SchmidhuberWilfried BrauerSepp HochreiterWolfgang Krull
The Roger Schank line2 on the timeline · 5 people
Roger SchankJacob L. MeyLouis HjelmslevRobert WilenskyPeter Norvig
The John Hopfield line2 on the timeline · 4 people
John HopfieldAlbert OverhauserCharles KittelTerry Sejnowski
The Marvin Minsky line2 on the timeline · 4 people
Marvin MinskyAlbert W. TuckerSolomon LefschetzPatrick Winston
The Shane Legg line2 on the timeline · 4 people
Harald FritzschMarcus HutterShane LeggJan Leike
The Yoshua Bengio line2 on the timeline · 4 people
Yoshua BengioRenato De MoriIan GoodfellowAaron C. Courville
What the Lines Show
Piaget
Papert’s teacher
Seymour Papert studied under the child psychologist Jean Piaget — which is why Logo was about how children think.
Simon
The CMU line
Herbert Simon taught Allen Newell and Edward Feigenbaum; between them they trained Richard Fikes and Douglas Lenat.
4 steps
Skinner to Abbeel
B. F. Skinner → William Estes → David Rumelhart → Michael I. Jordan → Andrew Ng: behaviourism to deep learning in four handovers.
No root
A forest, not a tree
The chains never meet. AI’s researchers descend from separate mathematical, psychological and linguistic traditions.
134 advisor links across 183 people · doctoral advisor data from Wikidata (CC0)Hover a name to trace its chain