A test of whether you can tell a machine from a person just by chatting with them.
The Turing test comes from the "imitation game" that British mathematician Alan Turing proposed in a 1950 paper. A judge chats in text with two hidden partners: one is a person, one is a machine. If the judge can't reliably tell which is the machine, the machine passes.
"Can machines think?" is hard to answer, so Turing swapped it for a question about behavior you can observe. Some recent experiments found chat AIs hard to tell apart from people in short conversations. Still, many argue that sounding human isn't the same as understanding, so today it's seen as one yardstick among many.
Think of it like this
Imagine chatting with a stranger in a group chat. Could you tell from the messages alone whether it's a classmate or a chatbot? Turing turned that guessing game into a test of machine intelligence.
For example
Turing replaced "Can machines think?" with "Can a machine do well at this game?"
Common mix-ups
Mix-upPassing the Turing test proves a machine thinks like a person.
ActuallyPicture someone who replies to letters in a language they don't know, using only a rule book to put together sensible answers. From outside it looks as if they understand, but they may not know the meaning at all — a well-known objection along these lines.
Mix-upTuring answered "Can machines think?" directly.
ActuallyIn his 1950 paper "Computing Machinery and Intelligence", after reframing the question, Turing spent many pages listing expected objections — such as "machines can never create anything new" — and examining them one by one.
Try it in class
For teachers at the start of a lesson — or for students working on their own.
- Question
- Could you tell from writing style alone whether you're chatting with a person or a machine? What clues would you use?
- Activity
- Play the imitation game with three roles. A judging group sends questions on paper slips; of two hidden groups, one answers itself and the other copies out answers written by an AI. After five rounds, the judges guess which is the machine and list their clues on the board. Finish with a debate: if the clues were well hidden, would that count as thinking?
Records about this
Source-checked moments from AI history.
Set up a framework for judging machine intelligence by behavior rather than by internal structure. The prototype of today's debates over benchmarks