Turing Test
A test proposed by Alan Turing in 1950 in which a machine passes if a human judge, chatting via text, cannot reliably tell whether responses come from the machine or another person.
The Turing Test, originally called the “imitation game,” was Alan Turing’s attempt to replace the vague question “can machines think?” with something measurable: can a machine’s text conversation be distinguished from a human’s?
The setup is simple. A judge exchanges messages with two unseen parties — one human, one machine. If the judge cannot reliably identify the machine, it passes. Turing predicted that by around 2000, machines would fool an average interrogator roughly 30% of the time in short conversations.
The test has always been controversial:
- Deception, not intelligence: it measures the ability to imitate, not understanding; a model trained on human text can mimic style without grasping meaning.
- Anthropomorphism: humans readily attribute minds to fluent text, a bias modern chatbots exploit.
- Cultural shift: judges may know the other party is an AI and still be convinced — or treat fluent, overconfident errors as telling.
Modern large language models routinely pass informal versions of the test, which is precisely why researchers now favour benchmarks that measure specific capabilities — reasoning, coding, planning — over a single pass/fail conversation.