
On this page
- AI history timeline at a glance
- How did AI start? (1943–1956)
- What was the golden age of AI? (1956–1974)
- What caused the first AI winter? (about 1974–1980)
- Why did expert systems boom and then bust? (1980–1993)
- How did AI come back? (1993–2011)
- What started the deep learning boom? (2012–2021)
- What happened after ChatGPT? (2022–2026)
- What does AI history teach us?
- Frequently asked questions
- Sources
The history of AI is a story of big promises, two crashes and one boom that hasn’t stopped. This timeline covers the origin of AI in 1950s Britain and New Hampshire, the AI winters, the deep-learning breakthrough and the generative-AI era up to 2026. For who founded the field, see who invented AI; for what AI means today, see what is artificial intelligence.
Short answer: AI history begins with Alan Turing’s 1950 question “Can machines think?” and the 1956 Dartmouth workshop, where the field got its name. Symbolic AI boomed until the first AI winter (about 1974–1980), expert systems boomed until the second (from 1987), and deep learning took over after 2012, leading to ChatGPT in 2022.
AI history timeline at a glance
| Year | Milestone | Why it mattered |
|---|---|---|
| 1943 | McCulloch and Pitts model the artificial neuron | First mathematical model of a neural network |
| 1950 | Turing publishes “Computing Machinery and Intelligence” | Proposes the imitation game (Turing test) |
| 1955–56 | Logic Theorist (Newell, Simon, Shaw) | Usually called the first AI program |
| 1956 | Dartmouth workshop | The field is named and founded |
| 1958 | Rosenblatt’s perceptron; McCarthy’s Lisp | First learning neural net hardware; AI’s main language for decades |
| 1964–66 | ELIZA (Joseph Weizenbaum, MIT) | First famous chatbot |
| 1969 | Minsky and Papert publish Perceptrons | Highlights limits of simple neural nets; research shifts to symbolic AI |
| 1973 | Lighthill report (UK) | Criticizes AI’s failure to deliver; UK funding cut |
| ~1974–1980 | First AI winter | Funding dries up in the UK and US |
| 1980s | Expert systems boom; Japan’s Fifth Generation project (1982) | First commercial AI wave |
| 1986 | Backpropagation popularized (Rumelhart, Hinton, Williams) | Lets multi-layer neural nets learn |
| ~1987–1993 | Second AI winter | Lisp-machine market collapses; expert systems prove brittle |
| 1997 | IBM Deep Blue beats Garry Kasparov | First defeat of a reigning chess world champion in a match |
| 2012 | AlexNet wins ImageNet | Starts the deep-learning boom |
| 2016 | AlphaGo beats Lee Sedol 4–1 | Go, long seen as too hard for machines, falls |
| 2017 | Transformer paper (“Attention Is All You Need”) | Architecture behind GPT, Claude and Gemini |
| 2022 | ChatGPT launches (30 Nov) | AI goes mainstream |
| 2024 | Nobel Prizes for Hopfield and Hinton (Physics) and Hassabis, Jumper and Baker (Chemistry); EU AI Act in force (1 Aug) | AI research reaches the Nobel stage; first comprehensive AI law |
| 2025–26 | Labs openly target AGI and superintelligence | See our superintelligence timeline |
Sources linked below. Winter dates are approximate; historians draw the edges differently.
How did AI start? (1943–1956)
Short answer: with neuroscience, logic and one famous question.
Warren McCulloch and Walter Pitts showed in 1943 that simple artificial neurons could compute logical functions. In 1950 Alan Turing asked whether machines can think and proposed testing them by conversation. In August 1955 John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon proposed a summer workshop on “artificial intelligence”, and in 1956 it met at Dartmouth College. Newell and Simon arrived with the Logic Theorist, a program that proved mathematical theorems.
What was the golden age of AI? (1956–1974)
Short answer: two decades of symbolic AI and big predictions.
Researchers built programs that solved algebra problems, played checkers and held simple conversations. Joseph Weizenbaum’s ELIZA (1964–66) imitated a psychotherapist with pattern matching, and users still confided in it, which alarmed its creator. Government money, mostly from DARPA in the US, flowed freely. Leading researchers predicted human-level machines within a generation. Meanwhile Frank Rosenblatt’s perceptron (1958) pointed toward learning machines, but Minsky and Papert’s 1969 book Perceptrons showed the limits of single-layer networks, and neural-net research faded for years.
What caused the first AI winter? (about 1974–1980)
Short answer: missed promises meeting tight budgets.
Programs that worked on toy problems failed on real ones because of the “combinatorial explosion”: the number of possibilities grew too fast for 1970s computers. In 1973 the Lighthill report, commissioned by the UK Science Research Council, concluded that AI had not delivered on its grand claims, and British funding was cut. In the US, DARPA moved toward narrower, mission-driven projects. “AI” became a label researchers avoided.
Why did expert systems boom and then bust? (1980–1993)
Short answer: they made money, then proved brittle and expensive.
Expert systems encoded the rules of human specialists. They found real commercial use, and in 1982 Japan launched its ambitious Fifth Generation Computer Systems project, which prompted rival programs in the US and UK. In 1986, Rumelhart, Hinton and Williams popularized backpropagation, reviving neural networks in research. But from about 1987 the market for specialized Lisp machines collapsed as cheaper desktop computers caught up, expert systems were costly to maintain, and the Fifth Generation project missed its goals. That was the second AI winter.
How did AI come back? (1993–2011)
Short answer: quietly, through statistics, data and faster chips.
AI rebranded as “machine learning” and focused on measurable tasks: spam filters, search ranking, speech recognition, recommendation. In May 1997 IBM’s Deep Blue beat world chess champion Garry Kasparov in a six-game match. The web produced huge datasets, and graphics chips (GPUs) made training large models practical.
What started the deep learning boom? (2012–2021)
Short answer: AlexNet in 2012.
Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton’s AlexNet won the 2012 ImageNet competition by a wide margin, proving deep neural networks trained on GPUs beat hand-built methods. Big tech companies hired the field’s researchers (the godfathers of AI among them). DeepMind’s AlphaGo beat Lee Sedol 4–1 in March 2016. In 2017 Google researchers published the transformer architecture, which scaled cleanly with more data and compute and became the basis for large language models.
What happened after ChatGPT? (2022–2026)
Short answer: AI went mainstream, and the goal shifted to AGI.
OpenAI released ChatGPT on 30 November 2022, and generative AI went from research demo to everyday tool within months. In 2024 the Nobel Prize in Physics went to John Hopfield and Geoffrey Hinton for neural-network foundations, and the Chemistry prize went partly to Demis Hassabis and John Jumper for AlphaFold. The EU AI Act entered into force on 1 August 2024. By 2025–26 the leading labs were openly aiming at artificial general intelligence and beyond, which is where this site’s main topic, super intelligence, comes in.
What does AI history teach us?
Short answer: hype cycles are real, but so is progress.
Each era overpromised: the founders expected a summer, the 1960s expected a generation, and the 1980s expected expert systems to run companies. Each bust still left working tools behind. The difference in the 2020s is that the money, users and capability gains are far larger than in any previous boom, and the old winters were caused partly by computers that were too slow. Whether today’s boom ends in a third winter or in something like the technological singularity is the open question. AI’s story has also been told on screen for a century; see our AI movies guide.
Enjoyed this? Get the SI Briefing
One short email a week: what changed on the road to super intelligence, and why it matters. No spam, unsubscribe anytime.
Frequently asked questions
What is the origin of AI?
AI as a research field began at the 1956 Dartmouth workshop, organized by John McCarthy, who coined the term in 1955. The idea goes back further: Alan Turing asked ‘Can machines think?’ in 1950, and McCulloch and Pitts described artificial neurons in 1943.
What were the AI winters?
Periods when funding and interest in AI collapsed after promises went unmet. The first ran from about 1974 to 1980, after the Lighthill report and DARPA cuts. The second began around 1987, when the market for specialized Lisp machines and expert systems collapsed, and lasted into the 1990s.
When did AI become popular?
AI had hype cycles in the late 1950s–60s and the 1980s, but mass public use began on 30 November 2022, when OpenAI released ChatGPT. It became one of the fastest-growing consumer apps ever.
What was the first AI program?
The Logic Theorist, written by Allen Newell, Herbert Simon and Cliff Shaw in 1955–56, is usually called the first AI program. It proved theorems from Principia Mathematica.
What are the main eras of AI history?
Roughly: foundations (1943–1956), the golden age of symbolic AI (1956–1974), the first AI winter (1974–1980), expert systems (1980–1987), the second winter (1987–1993), machine learning’s quiet rise (1993–2011), deep learning (2012–2021) and generative AI (2022 onward).
14 Sources
- Turing, A. M. (1950) Computing Machinery and Intelligence, Mind
- Dartmouth AI proposal (31 Aug 1955)
- History of artificial intelligence (Wikipedia)
- AI winter (Wikipedia)
- Lighthill report (Wikipedia)
- Fifth Generation Computer Systems (Wikipedia)
- IBM: Deep Blue
- AlexNet (Wikipedia)
- Google DeepMind: AlphaGo
- Vaswani et al. (2017) Attention Is All You Need
- ChatGPT (Wikipedia)
- Nobel Prize in Physics 2024
- Nobel Prize in Chemistry 2024
- European Commission: AI Act enters into force (1 Aug 2024)