What was the first AI?
The Logic Theorist is generally called the first AI program. Built in 1955-1956 by Allen Newell, Herbert Simon, and J.C. Shaw at RAND, it proved mathematical theorems from Principia Mathematica — including 38 of the first 52 theorems in Chapter 2 — and was demonstrated at the 1956 Dartmouth workshop.
Why — the first-principles explanation
To call something "the first AI" you need a rule for what counts. The rule most historians use: the program has to reason toward a goal it wasn't handed the answer to. Not calculate. Not follow a fixed recipe. Search.
The Logic Theorist clears that bar. Newell, Simon, and Shaw pointed it at Principia Mathematica, the famously dense logic textbook by Whitehead and Russell, and told it to prove the theorems. It didn't have the proofs stored. It had the axioms, the legal moves, and a strategy for guessing which move was worth trying — what we now call heuristic search. It proved 38 of the first 52 theorems in Chapter 2, and for one of them it found a proof shorter and cleverer than the one in the book. Simon told his class he and Newell had "invented a thinking machine." It was demonstrated at Dartmouth in summer 1956, which is why it's the one on the plaque.
The rule matters because there are earlier contenders that fail it or beat it depending on how you squint. In 1951, Christopher Strachey's checkers program and Dietrich Prinz's chess program ran on the Ferranti Mark 1 — earlier, but game-playing, which some historians count and some call mere search. Arthur Samuel's IBM checkers program (from 1952) actually learned from experience, which is arguably more advanced, but it arrived in stages rather than one dated release. Go back further and you hit the 1943 McCulloch-Pitts neuron, which is math, not a program, and Turing's 1948-1950 work, which is a design, not a machine.
So the honest ranking: first program built to imitate human reasoning — Logic Theorist. First AI-ish program to run at all — Strachey's checkers, 1951. First program that learned — Samuel's checkers. The Logic Theorist wins the title because it was aimed squarely at thinking, and because it had the good fortune to be demoed at the workshop that named the field. One more the public confuses with this: ELIZA, Joseph Weizenbaum's 1966 chatbot at MIT that imitated a therapist by reflecting your words back at you. It's the first famous conversational AI — ten years after the Logic Theorist, and not the first AI.
An example that makes it click
Imagine handing a kid a maze and saying "get to the middle." A calculator can't do it — it only answers what you ask. A recipe-follower can't do it — it only knows steps you already wrote down. But a kid who tries a path, hits a wall, backs up, and tries a smarter path next time? That's reasoning.
The Logic Theorist was the first program that did the maze. Its maze was a logic textbook. It knew the rules and the exit, not the route. It tried, backtracked, and got to the middle 38 times out of 52 — and once found a shortcut the textbook authors had missed.
Key facts
- The Logic Theorist was developed in 1955-1956 by Allen Newell, Herbert Simon, and J.C. Shaw of the RAND Corporation and Carnegie Tech.
- It proved 38 of the first 52 theorems in Chapter 2 of Whitehead and Russell's Principia Mathematica, and found one proof more elegant than the original.
- It was presented at the Dartmouth Summer Research Project on Artificial Intelligence in summer 1956.
- Christopher Strachey's checkers program and Dietrich Prinz's chess program ran on the Ferranti Mark 1 in 1951, predating it.
- Arthur Samuel's IBM checkers program, begun in 1952, was the first notable program to improve through learning.
- ELIZA, the first famous chatbot, was written by Joseph Weizenbaum at MIT in 1966 — a decade after the Logic Theorist.
▶ The 60-second explainer (script)
What was the first AI? The Logic Theorist, 1955 to '56, by Allen Newell, Herbert Simon, and J.C. Shaw. Here's why it gets the title. To count as AI, a program has to reason toward a goal it wasn't given the answer to. Not calculate — search. So these three pointed their program at Principia Mathematica, one of the hardest logic books ever written, and told it to prove the theorems. It didn't have the proofs. It had the axioms, the legal moves, and a strategy for guessing which move to try — that's called heuristic search, and that's the whole ballgame. It proved thirty-eight of the first fifty-two theorems in chapter two. And for one of them, it found a proof shorter and more elegant than the one Whitehead and Russell published. Simon walked into his class and said he'd invented a thinking machine. It got demoed at the Dartmouth workshop in the summer of 1956 — the same event that gave the field its name. Now, there are earlier contenders. Christopher Strachey had a computer playing checkers in Britain in 1951. Arthur Samuel had one at IBM learning checkers from 1952. And ELIZA, the famous fake therapist chatbot everyone thinks was first? That's 1966. Ten years late. The Logic Theorist wins because it was the first one built specifically to copy how humans think.
What authoritative sources say
People also ask
Was ELIZA the first AI?
No. ELIZA was written by Joseph Weizenbaum at MIT in 1966, ten years after the Logic Theorist. It was the first famous chatbot, which is why people remember it as first.
What was the first AI chatbot?
ELIZA, 1966. It imitated a psychotherapist by rephrasing whatever you typed as a question. Weizenbaum was disturbed by how many people confided in it despite knowing it was a simple script.
Was the Logic Theorist really the first?
It's the standard answer, but not undisputed. Christopher Strachey's 1951 checkers program ran earlier. The Logic Theorist gets the title because it targeted human-style reasoning rather than one game, and because it was demoed at the founding Dartmouth workshop.
What was the first AI robot?
Shakey the Robot, built at Stanford Research Institute from 1966 to 1972, is usually called the first mobile robot that could reason about its own actions rather than just execute commands.
What did the Logic Theorist actually do?
It proved theorems in symbolic logic by searching for valid proof paths, using rules of thumb to avoid trying every possibility. It handled 38 of the first 52 theorems in Chapter 2 of Principia Mathematica.