Turing Test

Proposed by Alan Turing in 1950, the Turing Test evaluates whether a machine's conversational responses are indistinguishable from a human's. A judge interacts via text with both a human and a machine; if the judge cannot reliably tell them apart, the machine is said to pass.

Modern LLMs have largely rendered the original Turing Test obsolete as a meaningful benchmark, spurring new evaluations focused on reasoning, factual accuracy, and alignment rather than mere human-likeness.