Chapter 2 of 8
The AI Timeline
The ideas behind AI are centuries old, the name is from 1955, and most people were using it every day long before November 2022 gave them a word for it.
Updated as AI changes. Last reviewed October 1, 2026.
Thirty-three moments across five eras, every one with its source attached, because a date without one is a story. Switch on Ryan’s own timeline to see where a practitioner was standing when each one happened.
“We’ve always thought about this neural network and this neural learning, and how does it all connect? What is the connective tissue? And what is it capable of?”
Foundations
1642–1950 6 moments-
1642
A machine that adds
Blaise Pascal begins his mechanical calculator, the Pascaline. The first step toward handing thinking work to a machine.
-
1837
The first computer design
Charles Babbage describes the Analytical Engine, a programmable machine that was never fully built in his lifetime.
-
1854
Logic becomes math
George Boole publishes The Laws of Thought. His true/false algebra is the base of every digital computer.
-
1936
The Turing machine
Alan Turing describes an imaginary machine that can compute anything computable. Modern computing starts here on paper.
-
1943
The artificial neuron
Warren McCulloch and Walter Pitts model a brain cell in math. Today’s neural networks descend from this idea.
-
1950
"Can machines think?"
Turing publishes "Computing Machinery and Intelligence" and proposes the imitation game, later called the Turing test.
Naming it
1950–1973 6 moments-
1952
A machine hears numbers
Bell Labs’ "Audrey" recognizes spoken digits. Voice assistants are 59 years away.
-
1955
The name is coined
John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon use the term "artificial intelligence" in a proposal dated Aug. 31, 1955.
-
1956
The Dartmouth workshop
The summer study proposed in 1955 meets at Dartmouth College. It is usually called the birth of AI as a field.
-
1962
IBM Shoebox
IBM shows a machine that understands 16 spoken words at the Seattle World’s Fair.
-
1966
The first chatbot
Joseph Weizenbaum’s ELIZA at MIT imitates a therapist by reflecting your words back. People confided in it anyway.
-
1973
The first AI winter
The Lighthill Report leads Britain to cut AI funding. Research money dries up on both sides of the Atlantic, roughly 1974 to 1980.
Hiding in plain sight
1980–2011 13 moments-
1980
Pac-Man’s ghosts
Each ghost follows its own chase rule, which is why they seem to have personalities. Early game AI.
-
1992
Fraud alerts
FICO’s Falcon uses a neural network to flag card swipes that don’t fit your pattern.
Machine learning fico.com : FICO Falcon, 1992 (opens in a new tab) -
1993
AutoCorrect
Microsoft Word starts fixing your typos as you type, from a built-in list.
-
1996
MapQuest directions
A shortest-path algorithm plans your route. Many people printed the directions and stopped memorizing roads.
-
1997
Deep Blue beats Kasparov
On May 11, 1997, IBM’s Deep Blue wins a six-game match against world chess champion Garry Kasparov.
-
1997
Clippy
Office 97 ships with the Office Assistant. It guessed what you were doing, often wrong.
-
1998
Google ranks the web
PageRank orders search results by how many important pages link to a page.
-
2000
Netflix recommends
Cinematch predicts how you will rate a movie from people with similar taste. In 2006 Netflix offers $1 million to anyone who can beat it.
-
2002
Roomba
iRobot launches a vacuum that navigates with simple reactive rules. No map, no learning.
-
2003
The spam filter learns
Microsoft’s SmartScreen filter learns what spam looks like from millions of user-labeled messages.
-
2006
Google Translate
Launches with statistical translation learned from huge amounts of already-translated text.
-
2011
Watson wins Jeopardy!
IBM Watson beats Ken Jennings and Brad Rutter in episodes aired Feb. 14–16, 2011.
-
2011
Siri
Siri ships built into the iPhone 4S in October 2011. You had to phrase things just right.
-
2010
The first pitch deck (early uCondition)
“When we first started, it was all about big data: machine learning, predictive analytics, and prescriptive analytics. If I could distill all those early pitch decks into that.”
The full story
The algorithm era
2012–2021 5 moments-
2016
Your feed stops being chronological
Twitter (February) and Instagram (March) switch to algorithm-ranked feeds. What you see is now chosen for you.
-
2016
Translation goes neural
Google Translate switches to neural machine translation.
-
2017
The transformer
Google researchers publish "Attention Is All You Need." The "T" in GPT comes from this paper.
-
2017
Face ID
The iPhone X unlocks by recognizing your face with a neural network.
Machine learning apple.com : Apple, iPhone X, 2017 (opens in a new tab) -
2018
The phone call that fooled people
Google demos Duplex calling a hair salon and a restaurant in a human-sounding voice.
Machine learning ai.googleblog.com : Google Duplex, May 2018 (opens in a new tab)
-
2013–2014
Beer Starts Here: the app, and the data they turned off
“It was all about using machine learning and predictive analytics to basically predict what beer consumers were most likely to drink in a given zip code. It weighed brands and it weighed styles of beer, ratings, and rankings.”
The full story -
2016–2018
Anutra Innovations: named for the initials
“Anutra was the parent company, the medical device company, and we were the spinout, but we named it Anutra Innovations because the initials were AI, and that was intentional from the beginning.”
The full story -
Same era
Predicting trade show sales
“We had crazy data inputs: what the weather was in the place where the trade show was, what day of the week it was, who the person selling was, what outfit we were wearing (company outfit or people wore their own stuff), company logo, polo tee, who else was in the booth, where the booth was. I mean, we had 30 or 40 data points, and then even more on the salespeople.”
The full story -
2018–2020
Boon: matching people in seconds
“I made the developers show me on the backend the algorithm weights and how they were self-adjusting and what modifications it was. I had manual overrides. We could see these things unfolding in real time.”
The full story -
2021–2023
Peoplelogic: predicting who leaves
“…at a key point in time where they were trying to determine the direction of the platform, and we really leaned into AI prediction.”
The full story -
Summer 2022
The demo, months before ChatGPT
“I’ll never forget, I was at Elon University in the summer of 2022. This was before ChatGPT was released.”
The full story
The generative era
2022– 3 moments-
Nov. 30, 2022
ChatGPT launches
OpenAI releases ChatGPT to the public as a free "research preview."
-
Jan. 2023
Fastest to 100 million
Analysts estimate ChatGPT reached about 100 million monthly users in two months (UBS, from Similarweb data).
-
2023–2026
From novelty to normal
Share of U.S. adults who say they have ever used ChatGPT: 18% in 2023, 23% in 2024 and 34% in 2025. Pew’s early 2026 reading is 44%, though that survey’s question also covered other chatbots, so it is not a like-for-like comparison.
-
Today
Teaching, researching, reading
“But my overall role in AI today is to stay aware of the trends and to think about the implications it has on human formation in the next generation.”
The full story
Two things you will not find above
The 1967 Senate prediction that Americans would work a 22-hour week and retire at 38. Only the retirement half is on the record, in Senator Walter Mondale’s July 26, 1967 opening statement to the Senate aging subcommittee. The 22-hour week has no primary source, so it stays a stage story and off this page.
The adoption-speed chart. How fast each technology reached a quarter and then half of the country is a different argument from this one, and it belongs with the framework built on it. It lives on the Velocity Gap page.
