5. History — The Inception of AI (1943–1956)

Source: AIMA 4th Ed, §1.3.1–1.3.2


Setting the Stage

AI as a formal discipline was born between 1943 and 1956. This was enabled by three converging forces: 1. A mathematical model of neurons (1943) 2. A formal theory of computation (Turing, 1936–1950) 3. The emergence of programmable digital computers (late 1940s)


Key Milestones

1943 — The First Neural Network Model

McCulloch & Pitts published “A Logical Calculus of the Ideas Immanent in Nervous Activity.” - Model: a neuron is a binary threshold unit — it fires (1) if the sum of inputs exceeds a threshold, otherwise does not (0). - Key result: a network of such units can compute any computable function. - This proved that a neural mechanism could in principle realize intelligence.

1949 — Hebbian Learning

Donald Hebb proposed his famous learning rule in The Organization of Behavior: > “When an axon of cell A is near enough to excite cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A’s efficiency… is increased.”

1950 — The Turing Test

Alan Turing published “Computing Machinery and Intelligence” in Mind. - Proposed the imitation game (Turing Test): a machine passes if a human interrogator cannot distinguish it from a human via text conversation. - Anticipated nearly every major objection to machine intelligence and refuted them. - Also proposed that rather than trying to simulate an adult mind, it is better to simulate a child’s mind and then educate it — the learning approach.

1951 — First Neural Network Program

Marvin Minsky and Dean Edmonds built the first neural network computer, SNARC (Stochastic Neural Analog Reinforcement Calculator) — 40 neurons, simulated on vacuum tubes and motor-driven clutches.

1956 — The Dartmouth Workshop (Birth of AI as a Field)

John McCarthy organized a 2-month summer workshop at Dartmouth College: - Attendees: McCarthy, Minsky, Shannon, Simon, Newell, Samuel, and others. - McCarthy coined the term “artificial intelligence” (rather than “cybernetics” or “automata theory”). - No consensus was reached, but the field had a name and a community.

Also 1956: Simon and Newell’s Logic Theorist was presented — a program that could prove theorems from Whitehead and Russell’s Principia Mathematica, including a proof of Theorem 2.85 that was shorter than Russell’s own proof.


Why This Period Matters

These years established three pillars that still hold: 1. Neural networks: intelligence can be modeled as networks of simple units. 2. Computation theory: what can/cannot be computed; the Church-Turing thesis. 3. Symbolic AI: formal logic and search can produce reasoning behavior.

The field would spend the next 70 years developing — and fighting over — these three pillars.