Artificial intelligence research began from two documents: Alan Turing's 1950 paper "Computing Machinery and Intelligence" and a 1955 proposal by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon for a summer study at Dartmouth College. Both rest on one bet, that human thinking can be described precisely enough for a machine to carry it out. The documents record research aims, not every motive of the people involved.
Key takeaways
- Turing's paper appeared in the journal Mind in October 1950 and replaced the question "Can machines think?" with a testable imitation game.
- The Dartmouth proposal is dated August 31, 1955, and asked for a two-month study by ten people in the summer of 1956.
- The proposal's central conjecture is that any feature of intelligence can be described so exactly that a machine can simulate it.
- The listed aims included machine language use, abstraction, problems reserved for humans, and self-improvement.
- Neither text proves that the bet is right, and the question is still argued.
Turing's 1950 question and the game that replaced it
A. M. Turing opened "Computing Machinery and Intelligence" in Mind, volume 59, issue 236, with the question whether machines can think. He then argued that the question is too vague to answer and swapped it for an imitation game. In the game, an interrogator types questions to two hidden players and tries to tell which is which. Turing asked whether a digital computer could take one player's part and be mistaken for a person about as often as a person would be.
The reason for the swap was practical. Arguing over the word "think" settles nothing, while a procedure with a measurable outcome can be run and repeated. Turing also made a forecast: with storage of about a billion units, he expected a machine, within about 50 years, to hold an average interrogator to no better than a 70 percent chance of a correct identification after five minutes of questioning. That is a prediction about one narrow test, not a claim that machines would match human minds.
Why the paper ends with a child machine
The last sections of the paper turn from testing to building. Turing suggested that programming an adult mind directly might be harder than building something like a child's mind and educating it. He pictured a machine with little fixed structure that improves through teaching, with reward and punishment signals shaping its behavior. Modern machine learning works on a similar principle, learning from examples and feedback instead of hand-written rules, although the paper does not describe anything like today's systems. The idea matters for the origin story because it names a second reason to build such machines: to find out how learning works.
The 1955 Dartmouth proposal
Five years later, a four-person group wrote the document that gave the field its name. The proposal, dated August 31, 1955, was written by McCarthy of Dartmouth, Minsky of Harvard, Rochester of IBM, and Shannon of Bell Telephone Laboratories. It asked for a study in the summer of 1956 that would last two months and involve ten people, and the title of the project already used the phrase artificial intelligence. A history of the event notes that McCarthy sought money for the seminar from the Rockefeller Foundation.
The conjecture on which the proposal rests says that every aspect of learning or any other feature of intelligence can in principle be described so precisely that a machine can be made to simulate it. The proposal then lists problems to work on. The first section deals with automatic computers and holds that the main obstacle is writing programs that use the available machines well, not building bigger hardware. The second asks how a computer can be made to use language, on the view that much human thinking may amount to manipulating words by rules. Summaries of the full document add aims such as forming abstractions and concepts, solving problems then reserved for people, and letting machines improve themselves. The AAAI reprint, published in AI Magazine in December 2006, is the standard place to read it.
Two origin documents side by side
The table sets out what each text says it is trying to do. It is a reading aid built from the two sources, and the last column describes only what the texts themselves state.
| Feature | Turing, 1950 | Dartmouth proposal, 1955 |
|---|---|---|
| Kind of document | Journal paper by one author | Funding proposal by four authors |
| Starting point | Can machines think? | Can intelligence be described exactly enough to simulate? |
| Method suggested | An imitation game to test behavior, then a machine that learns like a child | A two-month study of language, abstraction, learning, and self-improvement |
| Obstacle it names | The vagueness of the word "think" | Writing programs that use existing computers well |
| What it does not do | Claim that a machine had passed the test | Offer evidence for the conjecture before the study began |
What the origin documents do not settle
Both texts state a hypothesis and a plan. Neither shows that the hypothesis is true, and the argument over it has continued ever since: some researchers hold that imitating behavior is enough to speak of intelligence, others that behavior alone says little about understanding. The same split appears in current debates over whether language models reason. Readers should also be careful with phrases such as "the reasons behind the creation of AI". The texts give research reasons, namely curiosity about thought, a belief that computers were general enough to try, and an interest in learning machines. Funding pressures, military interest, and business goals are separate topics that these two documents do not cover, so they are not discussed here.
This article is general historical information. It makes no claim about the safety or capability of any current product.
FAQ
Who created artificial intelligence?
No single person did. Turing set out the question and a test in 1950, and the four authors of the Dartmouth proposal organized the 1956 project under the name artificial intelligence. Many other researchers built the first programs.
Did Turing think machines could think?
He called the original question too vague to discuss and proposed the imitation game as a stand-in. His prediction concerned how well a machine could play that game, not whether it had a mind.
Was the Dartmouth study a success?
It gave the field a name and a shared agenda. The plan called for a two-month study by ten people, a scale that fits a first exploration, and the field has since become a large research area.
Where can the original texts be read?
Turing's paper is available from the Oxford University Press page for Mind, and the Dartmouth proposal is available as a reprint from AAAI. Both links appear in the body of this article.
Related reading
- How AI developed after its beginnings
- The people behind early AI
- Whether AI can improve itself
- Telling real AI from hype
More in AI Policy and Ethics.
Last verified: October 2026
Sources
- Turing AM, Computing Machinery and Intelligence, Mind, volume 59, issue 236, October 1950
- McCarthy J, Minsky ML, Rochester N, Shannon CE, A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, August 31, 1955, reprinted in AI Magazine, volume 27, issue 4, AAAI, 2006
- Dartmouth Summer Research Project: The Birth of Artificial Intelligence, History of Data Science
- The Founding and Future of AI, Dartmouth College