liveBook · Manning

Turing test

As explained in Machines that Think

An evaluation in which an interrogator uses conversation to try to distinguish a machine from a human.

The Turing test evaluates whether a machine can imitate a human convincingly in conversation. Proposed by Alan Turing in 1950 as the imitation game, it replaces the vague question of whether machines think with an observable challenge: can an interrogator distinguish a machine from a person? section 1.3.4

An interrogator exchanges written messages with an unseen human and machine, then decides which is human. If the interrogator cannot reliably distinguish them, the machine passes. The test evaluates outward behavior—such as language, reasoning, and adaptability—rather than inspecting an internal process. section 1.3.4

Figure 1.5 The Turing Test. A human judge engages in written dialogue with two unseen participants. One is human and the other a machine, but their identities are hidden from the judge. If the judge cannot reliably distinguish the machine from the human based on their responses, the machine is said to pass the test.

The diagram shows the judge’s written exchanges with two participants whose identities are hidden. section 1.3.4

Passing is an operational result, not proof of consciousness or understanding. Fluent responses can create an impression of comprehension without settling what happens inside the system. As machines improve, judgments about what counts as intelligence also change. section 1.3.4 section 1.3.6

A five-minute conversation

In a 2025 Turing-test-style study, participants selected GPT-4.5 as the human 73% of the time when it was prompted to adopt a humanlike persona, more often than they selected the actual human. This result does not mean that all models are indistinguishable from people in every setting. section 8.5.2

Humanlike fluency also matters outside formal tests: conversational assistants can persuade, instruct, and entertain. The ability to blend into an exchange makes it important to distinguish persuasive appearance from evidence of understanding. section 1.3.6 section 8.5.2

How Machines that Think covers it

How deep it goes
Explains the written-dialogue setup and passing criterion, then examines why imitation does not settle questions of understanding.
Examples
Uses a five-minute GPT-4.5 conversation study to illustrate humanlike imitation and its contextual limits.
What you can do after
Evaluate and understand conversational imitation without treating fluency as proof of understanding.
How it connects
Connects behavioral intelligence to consciousness, understanding, and large language models.

Also explained in

Compare how 3 books explain Turing test →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machines that Think. The section links point to the parts of the book it draws on.