ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›ARTICLES›HISTORY OF AI: A TIMELINE OF KEY DATES
ARTICLES · EVERGREEN EXPLAINERBEGINNER

History of AI: a timeline of key dates from 1950 to today

From Turing’s question and two AI winters to deep learning and ChatGPT: the dates that mattered, what happened, and why each one changed the field.

READ10 min
WORDS1,956
SECTIONS7
TYPEEXPLAINER
CHECKED25 AUG 26
TL;DR — THE SHORT VERSION

The short version of more than seventy-five years of artificial intelligence.

  • The core ideas are old. Turing's test question, the perceptron and backpropagation all came long before today's models.
  • Scale is what changed the results. AlexNet showed that more data, more compute and bigger networks work, and the transformer was built in a way that scales.
  • ChatGPT changed access, not the science. The architecture already existed; what was new was a text box anyone could type into.
  • The field has misjudged its own timing in both directions. It was too optimistic early on and too dismissive later, so confident forecasts deserve caution.
  • The goalposts move. Chess, image recognition and conversation each counted as a test of intelligence until a machine did them.

The history of AI runs from Alan Turing’s 1950 question, “Can machines think?”, through two “AI winters” when money and faith ran out, to the deep-learning boom of the 2010s and ChatGPT in 2022. Most of what feels sudden about artificial intelligence was assembled slowly, by people who mostly did not expect this — and twice the whole field was written off as a dead end. Scroll to descend through the decades.

↓ SCROLL TO DESCEND

Key dates at a glance

The plain point: the whole timeline on one screen. Each date is explained and sourced in its section below.

YearEventWhy it mattered
1950Turing asks “Can machines think?”Swapped an unanswerable question for a test of behaviour.
1956Dartmouth workshop names the field; the Logic Theorist program appears“Artificial intelligence” becomes a field with a name and a mission.
1958Rosenblatt’s perceptronA machine that learns from examples: the ancestor of every neural network.
1970sThe first AI winterPromises went unmet and funding fell away.
1980sExpert systems boomAI as hand-written rules from human experts, sold to businesses.
1986Backpropagation goes mainstreamThe maths that lets many-layered networks learn.
Late 1980s–90sThe second AI winterExpert systems and their special hardware fell out of favour.
1997Deep Blue beats Kasparov at chessA machine beats a reigning world champion.
2011IBM Watson wins Jeopardy!A machine beats champions at quiz questions in everyday language.
2012AlexNet wins ImageNetDeep learning stops being fringe: scale works.
2016AlphaGo beats Lee Sedol at GoA game thought to be a decade out of reach falls.
2017The transformerThe design behind nearly every chatbot since.
2020GPT-3One huge language model doing new tasks from a few examples.
2022ChatGPT launchesAI reaches the public through a text box.

1950s–60s: the birth of AI

1950: Turing asks “Can machines think?”

Alan Turing publishes Computing Machinery and Intelligence and sidesteps the unanswerable question of whether machines can think.1Not read at source: A. M. Turing, “Computing Machinery and Intelligence”, Mind LIX(236), October 1950. Not read at source: the journal returns 403 to an automated request, so no link is given for a page this site did not open. Checked 16 Sep 2026. Instead he proposes a test (now called the Turing test): if a machine's replies are indistinguishable from a person's, on what grounds do you deny it?

More than seventy-five years later, that swap — from what a thing is to what it can be observed to do — still shapes every argument about AI, including the arguments on this site.

1956: The field gets its name at Dartmouth

A summer workshop at Dartmouth College gathers a handful of researchers around a proposal that machines could be made to simulate learning and intelligence.Not read at source: the Dartmouth Summer Research Project on Artificial Intelligence, proposed August 1955 by McCarthy, Minsky, Rochester and Shannon, held summer 1956. Checked 16 Sep 2026. The phrase "artificial intelligence" is coined largely to distinguish the work from existing fields; the European Commission’s research centre credits John McCarthy with it.1 The same year, Allen Newell, Herbert Simon and J. C. Shaw built the Logic Theorist, which Carnegie Mellon University calls the first artificially intelligent computer program.2

The proposal estimated significant progress in a couple of months. That optimism became a permanent feature of the field.

1958: A machine learns from examples

Frank Rosenblatt builds the perceptron, an early neural network:Not read at source: F. Rosenblatt, “The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain”, Psychological Review 65(6), 1958. Checked 16 Sep 2026. rather than being programmed with rules, it adjusts internal weights based on examples until it classifies them correctly. The press coverage promised far more than the device delivered.

The idea was right and a decade early. Every neural network since is a descendant.

1970s–90s: AI winters and expert systems

1970s: The first AI winter

The early promises did not arrive. Computers were too slow and too small for the problems researchers had said they would solve, and a British government review, the Lighthill report of 1973, concluded that nothing in the field had delivered the impact promised. Activity and funding dropped sharply: the first AI winter.1

It set the pattern for every hype cycle since: bold forecasts, then a backlash when they miss.

1980s: The expert systems boom

AI came back as expert systems: programs that stored a human specialist’s knowledge as a collection of facts and rules, plus an engine for applying them. Companies built them for real jobs, and the US, the UK and Japan put money into AI again.1

This was AI by rules, written down by people: the opposite of today’s learn-from-examples approach.

1986: Backpropagation goes mainstream

After a long stretch of disillusionment now called an AI winter, a paper popularises backpropagation3 — a method for assigning credit and blame through many layers of a network, so deep systems can actually learn.

This is the mathematical engine underneath everything that follows. The idea existed earlier; 1986 is when the field took it seriously.

Late 1980s–1990s: The second AI winter

Expert systems proved hard to build and keep up to date, because every rule had to be drawn out of busy human experts. By 1987 ordinary desktop computers had overtaken the expensive special machines built to run AI software. Through the 1990s the label fell so far out of favour that many researchers called their work something else: the second AI winter.1

1997: Deep Blue beats Kasparov

IBM's Deep Blue defeats world champion Garry Kasparov.Not read at source: the rematch ran 3–11 May 1997 in New York; Deep Blue won 3½–2½. Checked 16 Sep 2026.4 It is not learning in the modern sense — it is enormous, specialised search — but it lands as a cultural rupture.

It also sets a pattern that repeats to this day: a task is proof of intelligence until a machine does it, and afterwards it was "just calculation."

2010s: deep learning takes over

2011: IBM Watson wins Jeopardy!

IBM’s Watson beats the quiz show’s two most successful champions, Ken Jennings and Brad Rutter, on television. Unlike chess, the clues are written in everyday language, full of puns and wordplay, so the machine has to make sense of ordinary English.5

Another goalpost moved: answering trivia had looked like a human skill.

2012: Deep learning stops being fringe

A deep learning network called AlexNet wins the ImageNet image-recognition competition by a wide margin,6 using graphics processors to train at a scale that had not been practical before.

The lesson the field drew was blunt and consequential: scale works. More data, more compute, bigger networks. Everything after 2012 is partly a response to that finding.

2016: AlphaGo beats Lee Sedol at Go

DeepMind’s AlphaGo defeats Lee Sedol, one of the world’s best Go players, winning four of five games in Seoul.7 Go has so many possible positions that it was thought to be out of AI’s reach for at least another decade.1

Go had long been treated as a grand challenge for AI, so the result landed far beyond the game itself.1

2017: The transformer

A paper titled Attention Is All You Need introduces the transformer architecture,8 which processes a whole sequence at once and learns which parts to attend to, rather than reading strictly left to right.

It parallelises well, which means it scales — and scale had already been shown to work. Nearly every model you have heard of since is a transformer. See how one is trained.

2020s: generative AI

2020: GPT-3 shows what scale can do

OpenAI describes GPT-3, a language model with 175 billion adjustable numbers (parameters), ten times more than any comparable model before it, and tests how well it handles new tasks from just a few examples typed into the prompt.9

Two years later, the same family of models would sit behind a chat window.

2022: ChatGPT brings AI to the public

ChatGPT launches on 30 November 2022 and reaches a mass audience within weeks.10Not read at source (until 5 Oct 2026): the launch date is OpenAI’s own, widely recorded. An earlier version of this line said it reached an audience “faster than almost any product in history” — a superlative this site could not source, so it is narrowed to what is not in dispute. Checked 16 Sep 2026. Nothing fundamental was invented that day: the architecture was five years old and the model already existed. What changed was the interface — a text box anyone could type into. (The engine inside is a large language model.)

This is where the internet's composition begins to shift, and where the record on this site starts counting.

2026: Where we are standing

Frontier models ship at a pace measured in months rather than years, with an open-weights wave following close behind. Automated traffic has passed half the web, roughly half of new articles are machine-written, and the tools for telling the difference remain too weak to accuse anyone with.11

Read this stratum as unsettled. It is the one layer on this page still being poured.

THE MONOLITH
What more than seventy-five years of ideas eventually became.

What the pattern actually shows

Three things recur across more than seventy-five years, and they are more useful than any single date.

Progress is lumpy. Two long winters, then sudden jumps. People inside the field were repeatedly wrong about timing in both directions — too optimistic in the 1960s, too dismissive in the 1990s. Anyone confidently telling you what 2030 looks like is doing the same thing.

The goalposts move by design. Chess, then image recognition, then conversation: each was treated as the frontier of intelligence until it was crossed, then reclassified as mere mechanism. This is not dishonesty — it is what happens when you define intelligence by whatever machines cannot yet do.

The breakthroughs were mostly old ideas meeting enough compute. Neural networks date to the 1950s, backpropagation to the 1980s. What changed was hardware and data volume. That is worth remembering when a new capability appears to come from nowhere.

TAKEAWAY

Read any confident AI forecast against this record. The field has been wrong about its own pace in both directions, and a capability that seems to come from nowhere is often an old idea that finally met enough computing power.

◈ WHERE THIS SITE STANDS

This history is not a countdown to something. It is a record of a field that has been wrong about its own timeline in both directions for more than seventy-five years, and is still moving. The useful posture is neither the 1956 promise of imminent thinking machines nor the 1990s certainty that none of it would work — it is paying attention to what is measurable, dating it, and revising when the numbers move.

Questions people ask

Who invented AI?

No single person. Alan Turing asked the founding question in 1950, and John McCarthy coined the name “artificial intelligence” for the 1956 Dartmouth workshop that launched the field, with researchers such as Herbert Simon and Allen Newell among those attending.1

What was the first AI program?

The Logic Theorist, written in 1956 by Allen Newell, Herbert Simon and J. C. Shaw, is usually given that title: it proved theorems from a famous mathematics book. Carnegie Mellon University calls it the first artificially intelligent computer program.121

What was the AI winter?

An “AI winter” is a stretch when disappointment made money and interest in AI dry up. There were two: one in the 1970s, after early promises failed, and one from the late 1980s into the 1990s, when expert systems fell out of favour.1

When did ChatGPT come out?

On 30 November 2022, as a free “research preview”. OpenAI says it passed one million registered users by 5 December 2022.10

Related on this site: What Is an AI Winter? · What Is the Singularity in AI?

The “why it mattered” column of the key-dates table and the closing reflections are this site’s own summaries of the sourced sections, not claims from these sources.

SOURCES
  1. 1European Commission Joint Research Centre (Delipetrev, Tsinaraki & Kostić), AI Watch: Historical Evolution of Artificial Intelligence, EUR 30221 EN, 2020, read at source 5 Oct 2026“In 1950, Alan Turing published the milestone paper ‘Computing machinery and intelligence’ (Turing 1950), considering the fundamental question ‘Can machines think?’” ↑a“The first ‘AI period’ began with the Dartmouth conference in 1956, where AI got its name and mission. McCarthy coined the term ‘artificial intelligence,’ which became the name of the scientific field.” ↑b“The first AI winter started in the 1970s, due to unfulfilled promises, vast expectations, and financial difficulties.” … The Lighthill report (1973) “was pessimistic since ‘in no part of the AI field have discoveries made so far produced the major impact that was then promised.’” … “The first AI Winter has been marked by a dramatic decrease in the AI activities in both industry and academia.” ↑c“In the 1980s, the AI paradigm shifted to symbolic AI and the so-called ‘expert systems’ or ‘knowledge-based systems.’” … “Expert systems had two components: 1. the knowledge base – a collection of facts, rules, and relationships on a specific domain; and 2. the inference engine that described how to manipulate and combine these symbols.” It names the XCON expert system at Digital Equipment Corporation, the UK’s Alvey project and Japan’s fifth-generation computer project. ↑d“After the 1990s, the term ‘expert system’ dropped from the IT lexicon, and it was referred to as the second AI winter. One of the main problems in expert systems was knowledge acquisition.” … “by 1987, they had surpassed the expensive specialized Lisp machines produced by Symbolics and other manufacturers” … “Many AI researchers in the 1990s deliberately called their work by other names”. ↑e“Because of its complexity, the game of Go was considered out of AI reach for at least another decade.” ↑f“With more board configurations than atoms in the observable universe, the ancient Chinese game of Go has long been considered a grand challenge for AI.” ↑g“McCarthy coined the term ‘artificial intelligence,’ which became the name of the scientific field.” … “Among the conference attendees were Ray Solomonoff, Oliver Selfridge, Trenchard More, Arthur Samuel, Herbert A. Simon, and Allen Newell”. The 1955 Dartmouth proposal’s authors (McCarthy, Minsky, Rochester and Shannon) are noted, not read at source, in the 1956 section above. ↑h“The Logic Theorist had proven 38 theorems from Principia Mathematica”. ↑i“The first AI winter started in the 1970s, due to unfulfilled promises, vast expectations, and financial difficulties.” … “After the 1990s, the term ‘expert system’ dropped from the IT lexicon, and it was referred to as the second AI winter.” ↑j
  2. 2Carnegie Mellon University, History – The Simon Initiative, undated page, read at source 5 Oct 2026, under 1956“Simon, Newell and J. C. Shaw (CIT ’45,’46) of the Rand Corporation develop Logic Theorist, the first artificially intelligent computer program.” ↑
  3. 3Rumelhart, Hinton & Williams, “Learning representations by back-propagating errors”, Nature 323, 9 October 1986. Link checked 16 Sep 2026.↑
  4. 4IBM, Deep Blue, IBM Heritage, undated page, read at source 5 Oct 2026“Deep Blue again squared off against Kasparov in May 1997, this time at the Equitable Center in New York.” … “achieving a resounding victory, 3.5–2.5, in the rematch and becoming the first computer system to defeat a reigning world champion under standard tournament time controls.” ↑
  5. 5IBM, Watson, ‘Jeopardy!’ champion, IBM Heritage, undated page, read at source 5 Oct 2026“In a televised Jeopardy! contest viewed by millions in February 2011, IBM’s Watson DeepQA computer made history by defeating the TV quiz show’s two foremost all-time champions, Brad Rutter and Ken Jennings.” ↑
  6. 6Krizhevsky, Sutskever & Hinton, “ImageNet Classification with Deep Convolutional Neural Networks”, NeurIPS 2012. Link checked 16 Sep 2026.↑
  7. 7Google (DeepMind), What we learned in Seoul with AlphaGo, 16 Mar 2016, read at source 5 Oct 2026“DeepMind’s AlphaGo took on and defeated legendary Go player, Lee Sedol (9-dan professional with 18 world titles), marking a major milestone for artificial intelligence.” … “AlphaGo won four of the five games.” ↑
  8. 8Vaswani et al., arXiv:1706.03762, June 2017. Link checked 16 Sep 2026.↑
  9. 9Brown et al. (OpenAI), Language Models are Few-Shot Learners, arXiv:2005.14165, 28 May 2020, read at source 5 Oct 2026“we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting.” ↑
  10. 10OpenAI (Chatterji et al.), How People Use ChatGPT, 15 Sep 2025, read at source 5 Oct 2026“ChatGPT was released to the public on November 30, 2022 as a ‘research preview,’ and by December 5 it had more than one million registered users.” ↑a“ChatGPT was released to the public on November 30, 2022 as a ‘research preview,’ and by December 5 it had more than one million registered users.” ↑b
  11. 11Imperva / Thales, 2026 Bad Bot Report, read at source 17 Sep 2026automated traffic “accounting for more than 53% of all web traffic in 2025, up from 51% the year before” · Graphite, May 2026, read at source 17 Sep 2026: “since Q1 2025 the percentage of primarily AI-generated articles has plateaued at roughly 50%”. ↑
  12. 12Carnegie Mellon University, History – The Simon Initiative, undated page, read at source 5 Oct 2026“Simon, Newell and J. C. Shaw (CIT ’45,’46) of the Rand Corporation develop Logic Theorist, the first artificially intelligent computer program.” ↑

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK