Skip to content
BENIBIX Learn

Turing Test and Chinese Room

Level 4 · AI history

How the Turing test judges a machine by its chat, and why Searle said that passing is not the same as knowing. What ELIZA showed about us.

Course contentsLesson 7 of 10

After this lesson you can

  • Describe how the Turing test works
  • Explain the Chinese room argument
  • Say what the ELIZA effect is

12 mincovers 3 glossary entries

Worth reading first: Artificial Intelligence (AI)

Can a machine think, and how would we know?

The Turing test is a game for judging whether a machine can think. A judge chats by text with a person and a machine, and must say which is which. Alan Turing put it forward in 1950. He guessed that by about 2000, a judge would pick right no more than 70% of the time after five minutes.

Turing called it the imitation game. His trick was to swap a hard question for an easy one. Nobody can agree on what thinking is. But anyone can tell whether a chat feels human. If a machine fools the judge as often as a person would, Turing said, we have no fair reason to say it does not think.

In 1966, Joseph Weizenbaum at MIT wrote ELIZA, one of the first chatbots. It looked for key words and slotted them into set replies. Type 'I am sad about my job' and it might ask 'Why are you sad about your job?' It had no idea what a job was.

Even so, people opened up to it. Weizenbaum wrote that one of his staff asked him to leave the room so she could talk to it alone. Seeing a mind in a machine that only matches words is now called the ELIZA effect.

In 1980, John Searle replied with the Chinese room. A man who reads no Chinese sits in a room with a rule book. Notes in Chinese come in through a slot. He follows the rules to pick out the right signs and passes back perfect replies. People outside think he knows Chinese.

Searle's point was that the man only moves signs around, and so does a computer. Following rules for a language is not the same as knowing what it means. Many thinkers say he was wrong, and the debate is still open. It is one argument, not a proof.

In a 2025 study, judges picked a chat model as the human 73% of the time, more often than the real people. So the test has been passed in its simple form. That shows machines can sound like us. It does not settle whether they understand, which is just what Searle warned about.

The judge sees only text. If the machine is picked out no more often than chance, it passes.

Worked example

Emma plays the imitation game

Emma chats for five minutes with two hidden partners. One is her friend Jack, and one is a chatbot.

She plays ten rounds and spots the chatbot in only five. Guessing at random would score the same, so the chatbot passes her test.

The words for this lesson3

Check yourself

What does passing the Turing test show about a machine?

Finished this lesson?

Marking it complete also marks its 3 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.