Alan Turing’s 1950 proposal that instead of asking whether a machine can think, we should ask whether it can hold a conversation indistinguishable from a person’s. It shaped how the public thought about AI for seventy years, and current systems can pass versions of it while plainly not being what Turing’s readers imagined. The main lesson is that conversational fluency turned out to be a poor proxy for understanding.
Technical view
Formulated as the imitation game, with an interrogator distinguishing a machine from a human through text alone. It was criticised from the start as behavioural rather than explanatory, most famously by Searle’s Chinese Room argument, and as easily gamed by evasion and personality rather than competence. Controlled studies have reported modern models being judged human more often than actual humans in short exchanges. Practical evaluation has long since moved to capability benchmarks and task-specific assessment, though the test remains a useful reminder that fluency and reliability are independent.
Related
Part of AI Terminology Index, section: Adjacent Fields.