Imagine a computer that understands Chinese. It takes Chinese characters as input and produces Chinese characters as output. Suppose the computer program was so good that it passes the Turing test, meaning that it has convinced human Chinese speakers that it itself is an intelligent human being. Proponents of artificial intelligence would say that the computer understands Chinese in just the same way as a human Chinese-speaker does.

However, imagine we replace our computer with a room. The room has two slits like a post box as its input and output channels. The room also contains an English version of the same computer program used before, so that an English-speaking human could follow the same rules for processing Chinese characters. Imagine further that you are inside this room. You do not understand any Chinese but when a written message is pushed through the room's input channel, you are able to follow the instructions in the manual and write out an output message. You have no idea what the input or output mean but are able to produce meaningful answers. As the original computer was able to pass the Turing test, you too should be able to prove that you are an intelligent Chinese-speaking human.

However, in spite of your intelligence you had no understanding of the meaning of those Chinese messages. All you were doing was manipulating what to you were meaningless symbols. This thought experiment was used to show that artificial intelligence could never have any true understanding of what they were doing. Without understanding, how can computers be thought of as having minds?

Thought experiment courtesy of John Searle's "Minds, brains and programs" (1980)