The 1956 Dartmouth Workshop: How AI Got Its Name

A conjecture, not a finding

John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon signed the proposal on August 31, 1955, asking the Rockefeller Foundation for $13,500 to fund what they called a two month, ten man study of artificial intelligence at Dartmouth College the following summer. The study was to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. [2]

Atlas interpretation: That sentence is a hypothesis written in the grammar of a certainty. It is also the proposal's only real claim: nothing in the document says intelligence has been so described, only that it can be, in principle. The workshop was funded to test a premise its own authors had not yet established. [2]

Seven problems and one new name

The proposal listed seven areas for the group to attack: getting computers to use language, arranging hypothetical neurons into concepts, estimating the complexity of a calculation, getting machines to improve themselves, classifying and forming abstractions, and using controlled randomness to model creative rather than routine thought. "Artificial intelligence" was the proposal's own name for the combined subject, not a label attached afterward. [2]

Ten invited, three who stayed

The workshop ran from June 18 to August 17, 1956, roughly eight weeks, on the top floor of the Dartmouth math building. Attendance did not resemble ten people sharing a summer. Only Minsky, McCarthy and Ray Solomonoff stayed the entire time; everyone else, including Allen Newell and Herbert Simon, came and went as their own schedules allowed, some for a few weeks and some for a few days. The Rockefeller Foundation had also funded only about half of what the proposal requested. [3][4]

What the summer actually produced

The concrete work shown at Dartmouth had mostly been built before anyone arrived. Newell and Simon, present for only the opening weeks, demonstrated the Logic Theory Machine, a program they had already written that proved theorems from Whitehead and Russell's Principia Mathematica, and compared its output against human subjects' reasoning. Minsky presented early ideas toward a plane geometry theorem prover, and Alex Bernstein of IBM presented a chess program, which prompted McCarthy toward the alpha-beta pruning idea. [3]

Atlas interpretation: The founding overconfidence runs in two directions at once. The proposal assumed intelligence was already precisely describable in principle; the myth that grew up around the workshop assumes a group of ten spent a unified summer confirming it. Neither happened. What the summer actually left behind was a name, a list of problems, and a handful of people, rarely in the same room, comparing programs they had mostly built somewhere else. [2][3]