Turing Test
Alan Turing proposes the "Imitation Game", later known as the Turing Test, as a criterion for machine intelligence.
Overview
Alan Turing proposed the famous "Imitation Game" in his 1950 paper "Computing Machinery and Intelligence". The core idea of the test is that if a machine can converse with a human without being identified as a machine, then it can be said to have intelligence. The Turing Test remains an important concept in the field of artificial intelligence, though it has also sparked many philosophical debates.
Study Notes
Background, mechanics, impact, and key takeaways in one read
Background & Context
In 1950 Alan Turing published "Computing Machinery and Intelligence" in the journal Mind (vol. 59, no. 236). The question "Can machines think?" was mired in definitional confusion, and Turing argued that debating the meaning of "think" was fruitless. He replaced it with an operational alternative—the "imitation game": if a human interrogator cannot tell, from text conversation alone, whether they are speaking to a machine or a person, the machine may be said to act intelligently. The paper also anticipated ideas of machine learning and search, and systematically answered objections from mathematics, consciousness, and theology.
Key Points
- 1The paper replaces the ill-defined question "Can machines think?" with the operational "imitation game".
- 2The original game has three players: an interrogator distinguishes a man (A) from a woman (B) by text only.
- 3Turing puts a machine in A's role; if the interrogator cannot tell, the machine passes.
- 4He imagines a "child machine" that learns like a child and is then educated—an early vision of ML.
- 5The essay rebuts objections from Gödel's theorem, consciousness, and theology against machine intelligence.
- 6Turing predicted that by century's end people would speak of machines thinking without controversy.
How It Works
The imitation game is essentially a blind test. An interrogator (C) is separated from two respondents (A and B) and communicates only by typing, trying to decide which is the man and which the woman. Turing first framed the question as this sex-identification game, then substituted a machine for A. If the machine disguises itself well enough that the interrogator's error rate matches the human-versus-human case, it displays indistinguishable intelligent behavior. Turing also characterized a digital computer by three parts—store, executive unit, and control—and noted that any discrete-state machine can be simulated by a general-purpose computer. In other words, if a mental operation can be described by an instruction table, a machine can in principle reproduce it, moving "intelligence" from metaphysics to engineering.
Impact & Legacy
The Turing Test long served as a benchmark for machine intelligence, inspiring dialogue systems and chatbots, and provoking counterarguments such as Searle's Chinese Room. Its limitation is that fluent text need not imply real understanding, and it sidesteps consciousness. Yet Turing's shift—defining intelligence by observable behavior rather than inner awareness—gave AI a concrete, pursuable goal whose influence persists today.
Key Takeaways
- Framing a problem operationally—via a stand-in—can be more powerful than defining it.
- Intelligence can be defined by observable behavior, not inner consciousness—AI's engineering start.
- Turing's "child machine" foreshadowed learning from data rather than hand-coded rules.
Glossary
Turing's thought experiment, later called the Turing Test: judging machine intelligence by whether text dialogue is indistinguishable from a human's.
Searle's counterargument that symbol manipulation is not real understanding, aimed at the Turing Test.
Turing's machine that can simulate any discrete-state machine, supporting the claim that machines can perform any mental operation.