Definition
The Dartmouth Summer Research Project of 1956 is widely seen as the founding event of AI. John McCarthy, Marvin Minsky, Claude Shannon and Nathaniel Rochester proposed it in 1955, in the document that first used the term "artificial intelligence". Early programs that followed included the Logic Theorist, the General Problem Solver, ELIZA and the robot Shakey.
In simple words
A few researchers met one summer, sure that a small group could make big progress quickly. They underestimated how hard it was, but that meeting gave the field its name and its first community.
Why it matters
It shows AI's recurring pattern of great hope followed by hard reality. ELIZA also taught an early lesson: people readily believe a simple program understands them, and that still matters with chatbots.
Key terms
- Dartmouth workshop
- The 1956 summer meeting at Dartmouth College in the USA that founded AI as a field of research.
- John McCarthy
- The computer scientist who organised the workshop and coined the term "artificial intelligence".
- Logic Theorist
- A 1956 program by Allen Newell, Herbert Simon and Cliff Shaw that proved mathematical theorems. It is often called the first AI program.
- Symbolic AI
- The early approach to AI that represents knowledge as symbols and rules, and reasons with logic.
- Perceptron
- An early learning machine built by Frank Rosenblatt in 1958, the ancestor of today's neural networks.
Explained in detail
The proposal
In August 1955 four researchers wrote a short proposal asking the Rockefeller Foundation to pay for a summer of research. They were John McCarthy of Dartmouth College, Marvin Minsky of Harvard, Nathaniel Rochester of IBM and Claude Shannon of Bell Laboratories, who had founded information theory. The proposal introduced a new name for the subject, "artificial intelligence", and stated its central belief: 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.
They listed the problems they hoped to work on: how to make machines use language, form ideas, solve the kinds of problems then left to people, and improve themselves. They wrote that a carefully chosen group of scientists could make a significant advance if they worked on it together for a summer.
The summer of 1956
The workshop ran for about two months in the summer of 1956. It was less tidy than planned: people came and went, and no single breakthrough was made. Its importance lay elsewhere. It gave the field its name, and it brought together the people who would lead AI research for the next thirty years. Two of the visitors, Allen Newell and Herbert Simon, arrived with something remarkable: a working program called the Logic Theorist, which could prove theorems from a famous book on the foundations of mathematics. It eventually proved 38 of the first 52 theorems in that book, and found a shorter proof for one of them.
The first wave of programs
The years that followed were full of firsts. In the 1950s Arthur Samuel at IBM wrote a program that played draughts and improved by playing against itself; in 1959 he popularised the phrase "machine learning". In 1958 John McCarthy created LISP, a programming language used in AI for decades, and Frank Rosenblatt demonstrated the Perceptron, a machine that learned to recognise simple patterns. In 1966 Joseph Weizenbaum's ELIZA held typed conversations. From 1966 to 1972, a robot called Shakey at the Stanford Research Institute became the first mobile robot to perceive its surroundings, plan a route and carry out its plan.
Great confidence, and its cost
The early researchers were brilliant and far too optimistic. In 1958 Herbert Simon and Allen Newell predicted that within ten years a computer would be the world chess champion and would discover an important new mathematical theorem. The chess prediction came true thirty-nine years later, in 1997. Problems that looked simple, such as understanding a child's story or recognising a cup on a table, turned out to need enormous amounts of knowledge about the world. The gap between what was promised and what was delivered led to the first "AI winter", which is the subject of the next topic.
| Year | Milestone |
|---|---|
| 1950 | Turing asks "Can machines think?" |
| 1955 | The Dartmouth proposal coins "artificial intelligence" |
| 1956 | The Dartmouth workshop; the Logic Theorist |
| 1958 | LISP and the Perceptron |
| 1966 | ELIZA; work begins on Shakey the robot |
How it works
| Program | Year | Did |
|---|---|---|
| Logic Theorist | 1956 | Proved theorems from a mathematics textbook |
| General Problem Solver | 1957 | Tried a general method for solving problems |
| ELIZA | 1966 | Imitated a therapist using simple pattern matching |
| Shakey | 1966–72 | A mobile robot that planned its own actions |
Try it yourself
The road to the first AI meeting
Travel through the early years of the idea.
Place the early milestones in order
Arrange these events from earliest to latest.
- 1Shakey becomes the first robot to plan its own actions
- 2Joseph Weizenbaum's ELIZA chats like a therapist
- 3Alan Turing asks "Can machines think?" and proposes the imitation game
- 4Frank Rosenblatt builds the Perceptron, an early learning machine
- 5The Dartmouth workshop gives the field the name artificial intelligence
Common mistakes
- Thinking early AI failed entirely. Search, logic and planning from that era are still in use.
- Assuming that a program which sounds caring understands you. That's known as the ELIZA effect.
Check yourself
Summary
- Dartmouth, 1956, gave AI its name and start.
- Early programs worked in small worlds.
- ELIZA showed how easily people assume understanding.
Finished reading?
Sources: Dartmouth proposal (1955) · Russell & Norvig, Artificial Intelligence: A Modern Approach, ch. 1