Searle’s Chinese Room
Searle imagines what it would be like to operate as a computer performing tasks in a manner suggestive of intelligence. Can he find a counterexample where he, acting as a nonintelligent computer, passes the Turing test? In the Chinese room, Searle thinks of himself as following a program like a computer. The program (a set of rules) is designed to produce answers to questions formulated in Chinese, even though Searle understands no Chinese. He is locked in a room and receives paper slips with questions through a slot. When Searle receives a question, he follows the rules to produce an answer in Chinese.
Suppose that Searle’s answers are as good as those of a Chinese person. Searle would pass the Turing test for understanding Chinese. But if Searle doesn’t understand Chinese, how could a computer, doing the same thing? Sheer symbol shuffling leads to understanding neither in a human nor in a machine.
The Turing test concentrates on behavior but neglects the mind. As Searle points out, if the idea is that behavior is all we need to ascribe intelligence, then radios and TV sets would be intelligent as well. The Chinese room argument was designed to show that strong AI is false—intelligence doesn’t consist of formal symbol manipulation. The quest to program minds through syntax is hopeless because no matter how much syntax there is in a program, it cannot rise to the level of semantics. The problem is that syntax is not self-interpreting.
Searle’s thought experiment has been interpreted as an attempt to prove that machines couldn’t be conscious. Searle thinks, however, that we are conscious machines—ones with physical structures that causally sustain consciousness as a higher-level feature of the brain. Humans are conscious biological machines, and we might build artificial ones in the future. If so, we will build them out of physical components that are adequate for causally sustaining consciousness. They would need to have minimum causal powers for consciousness. But what could intelligence be, if not formal symbol manipulation? Let us explore an alternative view proposed by Searle.