Who created AI and when: the history from 1950 to ChatGPT
Who created AI? No one alone: the name was born in 1955, the field in 1956, and the AI you use in 2017. See the history, from pioneers and winters to ChatGPT.
NinaKalkazu.org
Listen to this article
Narrated by Thalita
For many people, artificial intelligence appeared in November 2022, the day ChatGPT started answering questions on their phone screens. But when you ask who created AI and when, the answer goes back more than 70 years and doesn't fit in a single name.
It was conceived, named, and built by different people at different times. Here you'll see the three answers to "when was AI created," who was behind each one, and why the story took so long to reach your pocket.
Quick summary: the answer in 30 seconds. AI has no single creator. The idea took shape in 1950, with mathematician Alan Turing. The name "artificial intelligence" appeared in 1955, in a proposal signed by John McCarthy and three other scientists, and the research field was born in the summer of 1956, at Dartmouth, in the United States. The AI you use today comes from 2017, and ChatGPT arrived in 2022.
Index
- Who created AI: the short answer
- 1950: Turing asks if machines can think
- 1955 and 1956: the name and the Dartmouth summer
- Why AI took so long: the two schools and the winters
- When the AI you use today was born
- Frequently asked questions
Who created AI: the short answer
No one created AI alone. The most cited name is John McCarthy, who named the field in 1955 and organized the 1956 meeting that turned it into a science. But the question "when" has three correct answers, depending on what you call AI: the idea, the research field, or the technology behind ChatGPT.
Think of aviation. There were those who dreamed of flying machines, those who named aeronautical engineering, and those who built the first plane that carried passengers. With AI it was the same, and each stage has its date.
| What is meant by "creating AI" | When | Who |
|---|---|---|
| The idea of a thinking machine | 1950 | Alan Turing |
| The name and research field | 1955 and 1956 | McCarthy, Minsky, Rochester and Shannon |
| The technology behind ChatGPT | 2017 | Google researchers |
| The AI that reached the public | 2022 | OpenAI, with ChatGPT |
1950: Turing asks if machines can think
The modern idea of artificial intelligence begins in October 1950, when British mathematician Alan Turing published the article "Computing Machinery and Intelligence" in the academic journal Mind. The first sentence already stated its purpose: "I propose to consider the question: can machines think?"
Turing knew that "thinking" is hard to define. So he replaced the question with a game, which he called the imitation game: a person converses in writing with a machine and with another human being, without seeing either. If they can't tell which is which, the machine passed the test.
This game became known as the Turing test. It measures intelligence by behavior, like a teacher who evaluates a student by the exam, without seeing what goes on in their head. The idea is debated to this day, as we show in the article about whether artificial intelligence can think like humans.
Turing's original text is still available on the journal publisher's website. He didn't build an AI: he gave the field the question it still pursues.
1955 and 1956: the name and the Dartmouth summer
The name "artificial intelligence" appeared in a document dated August 31, 1955: a research proposal signed by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon. They requested funding to bring together ten scientists for two months at Dartmouth College, in the United States, in the summer of 1956. This meeting is considered the birth of AI as a science.
The four came from different places. According to the original proposal, McCarthy taught at Dartmouth, Minsky was at Harvard, Rochester worked at IBM, and Shannon at Bell Laboratories.
Their bet was bold. The text said that every aspect of learning and intelligence could, in principle, be described so precisely that a machine could imitate it. Dartmouth itself still cites this sentence when telling the story of the meeting.
The name was almost different. The MIT Technology Review recalls that Shannon preferred "automata studies" and that a colleague complained that the word "artificial" sounded like something fake. McCarthy won, and the term stuck.
Why AI took so long: the two schools and the winters
Because the pioneers promised too much with weak computers and little data. The field split between those who wanted to teach rules to machines and those who wanted them to learn on their own. Both bets stalled, funding disappeared twice, and these periods of abandonment earned a name: AI winters.
The first school, the rules-based one, tried to write the logic of reasoning by hand, like a giant instruction manual. The second imitated the brain, with networks of artificial neurons that learn from examples, like a child who sees a thousand cats until recognizing one.
The second school had its great moment in July 1958. Psychologist Frank Rosenblatt presented the perceptron, a machine that learned on its own. Running on an IBM 704 computer weighing about five tons, it learned to tell whether a mark printed on a card was on the left or right side after 50 attempts, Cornell University reports.
Optimism was enormous. In 1967, Minsky wrote that the problem of creating artificial intelligence would be solved "within a generation," IEEE Spectrum recalls. It wasn't.
The AI winters
In the 1970s, results didn't match promises, and governments cut funding. A British government report concluded that the AI dream had gone nowhere, and many scientists began avoiding the term in funding requests, according to MIT Technology Review.
The 1980s brought a respite with expert systems, programs full of rules that imitated a doctor or lawyer in a narrow subject. The market collapsed at the end of the decade, in the second winter. In the 1990s, working with AI was out of fashion.
When the AI you use today was born
The AI behind ChatGPT, Gemini, and Claude was born in 2017 and reached the public in 2022. It depends on three ingredients that only came together this century: data in quantity, with the internet; fast chips, with graphics cards; and a new way to build neural networks. It was the school that imitated the brain, abandoned in the winters, that won.
The first sign came in 2012. Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton, from the University of Toronto, entered a neural network called AlexNet in a competition that asked to recognize objects from a thousand categories in photos. According to IEEE Spectrum, it was wrong 15% of the time, versus 26% for the runner-up. The difference woke up the industry.
The secret was in the chip. The network was trained on GPUs, graphics card processors, which do thousands of small calculations at once, like a market with hundreds of open checkout lanes. In 2006, Nvidia launched CUDA, the program that freed these chips for calculations beyond games.
The decisive step came in June 2017. Eight researchers, almost all from Google, published the paper "Attention Is All You Need", which presented the transformer: a way to build the network that reads the entire sentence at once, like someone looking at the whole page, not word by word. It's the "T" in GPT, the acronym for ChatGPT's models, and is the basis of almost all assistants today.
On November 30, 2022, OpenAI launched ChatGPT as a free research preview, built on GPT-3.5, the "brain" OpenAI had already trained. The company expected little. "We didn't want to sell it as a big breakthrough," one of those responsible told MIT Technology Review. In a short time, millions of people were using it.
Artificial intelligence timeline
| Year | Milestone |
|---|---|
| 1950 | Turing publishes "Computing Machinery and Intelligence" and proposes the imitation game |
| 1955 | McCarthy, Minsky, Rochester and Shannon sign the proposal that creates the name |
| 1956 | Summer research at Dartmouth: AI becomes a scientific field |
| 1958 | Rosenblatt presents the perceptron |
| 1970s | First AI winter |
| Late 1980s | Second winter, with the fall of expert systems |
| 2006 | Nvidia launches CUDA and opens GPUs to science |
| 2012 | AlexNet wins the image competition by a wide margin |
| 2017 | Google publishes the transformer |
| 2022 | OpenAI launches ChatGPT |
Frequently asked questions
Who is considered the father of artificial intelligence?
The title is usually given to John McCarthy, who created the name and organized the Dartmouth meeting. Alan Turing is also called that, for having formulated the question that gave rise to the field. The two answer different questions: Turing gave the idea; McCarthy, the name and the field.
What was the first artificial intelligence?
There is no consensus, because it depends on what is accepted as AI. The 1958 perceptron is one of the first examples of a machine that learned from practice, and is the direct ancestor of today's neural networks.
Was ChatGPT the first AI?
No. It arrived more than 60 years after Dartmouth. Before it, AI already filtered spam, recommended movies, and recognized faces in photos. ChatGPT was the first conversational AI to catch on with millions of people at the same time.
Why do people talk about "AI winter"?
It's the name given to periods when the field lost money and prestige for not delivering on its promises: one in the 1970s and another in the late 1980s. Both ended when more powerful computers and new techniques emerged.
AI took 72 years to go from Turing's question to your phone screen. Next time someone says it was "just invented," you already know the right date for each part of the story.
