Can a machine that speaks perfect Chinese understand a single word of it?
Picture a man locked in a room, following an English rulebook to answer Chinese notes slipped under the door. His replies are flawless, yet he understands none of it. In 1980 the philosopher John Searle argued that a computer running a program is in exactly the same position.
John Searle set out the Chinese room in a 1980 paper, 'Minds, Brains, and Programs', in the journal Behavioral and Brain Sciences. It became the journal's most influential target article and has drawn a huge number of replies. His target was what he called strong AI: the claim that a suitably programmed computer with the right inputs and outputs would have a mind in the same sense that people do.
The thought experiment runs like this. Suppose a program lets a computer converse in Chinese so well that no one can tell it from a native speaker. Now Searle himself sits in a room with an English version of that program, plus paper, pencils and filing cabinets. He follows the rules by hand, matching symbols to symbols, and pushes out perfect Chinese answers. He would pass the same test, yet he understands no Chinese. Since the computer does nothing more than he does, Searle concludes that running a program is not enough for understanding.
The idea had ancestors. In 1713 Leibniz imagined a brain enlarged to the size of a mill and found it hard to see how perception could arise from mechanical parts alone. In 1961 the Soviet writer Anatoly Dneprov published a story in which a stadium full of people act as the parts of a computer, translating Portuguese that none of them speak.
The most common objection is the systems reply: the man does not understand Chinese, but the whole system, man plus rulebook plus files, does. Searle answered that this simply assumes what it needs to prove, and that a pile of paper and a person cannot obviously become a new mind. Others argue for giving the program a robot body, or simulating a brain neuron by neuron. Note that the argument does not say machines cannot behave intelligently; it concerns whether behaviour alone amounts to understanding. It remains fiercely contested, and newly relevant in an age of fluent machines.
Source: Wikipedia — Chinese room · Text summarised from Wikipedia (CC BY-SA 4.0)
