Foundations
Turing Test
Also known as: imitation game
A classic thought experiment: if a machine can converse so that people cannot tell it from a human, does it "think"? Historic, but no longer a serious benchmark.
The Turing test is a famous old idea from 1950: if you chat with a hidden partner and cannot tell whether it is a human or a machine, then, the argument goes, the machine is behaving intelligently. For decades it was the symbolic goal of AI. Today's chatbots can often pass casual versions of it, but that turned out to reveal a flaw in the test more than proof of true understanding.
The Turing test, proposed by Alan Turing in 1950 (he called it the "imitation game"), sidesteps the question "can machines think?" by asking a practical one: can a machine converse well enough that a human judge cannot reliably tell it from another human?
Its historical role
For much of AI's history it was the iconic milestone. Modern LLMs can fool people in short conversations, effectively passing casual versions.
Why it is no longer a serious benchmark
Passing it shows convincing imitation, not genuine understanding, reasoning or truthfulness (models still hallucinate). The field now uses concrete, task-based benchmarks instead. The Turing test remains important historically and philosophically, not as a measure of real capability.
Turing's 1950 imitation game reframed "can machines think?" as an operational, behaviourist criterion, indistinguishability in conversation, deliberately avoiding definitional debates about thought. It profoundly shaped AI's framing but was always contested (e.g. Searle's Chinese Room challenges whether behavioural indistinguishability implies understanding).
Empirically, capable LLMs now pass many informal and some controlled versions, which underscores the test's core limitation: it measures persuasive human-imitation, a property LLMs optimise almost incidentally, rather than reasoning, factuality or robust competence. Consequently it is obsolete as a capability metric, superseded by targeted benchmarks and evaluations, though it retains historical and philosophical significance and informs adjacent concerns (deception, anthropomorphism, and how to define/measure progress toward general intelligence).