Chinese Room Properly Translating the Chinese Room

Chinese Room

Properly Translating the Chinese Room

John Searle's thought experiment concerning the "Chinese Room" attempts to disprove that so-called "strong-AI" (artificial intelligence that demonstrates "true" thinking and "understanding") could ever possibly exist. The argument is relatively straightforward: Searle imagines a computer running a program that allows it to communicate in written Chinese -- the program is capable of recognizing Chinese characters that are entered into it and of formulating a response in written Chinese that makes sense and appears conversant. This program is so perfect that it can convince a native Chinese speaker that the responses, generated solely through a series of complex rules or algorithms that the computer/program uses to react to the Chinese characters it receives as input, are coming from an actual human Chinese speaker. Next, Searle imagines that if he were placed in a room with this computer and if written Chinese communications were slipped under the door, he would be able (through use of the computer) to deliver appropriate responses to these communications without ever understanding a word of Chinese. As the human respondent (in this case Searle) is delivering Chinese responses without any understanding, he concludes that the computer program is working similarly, generating automatic if complex responses without actually understanding the language.

While Searle's though experiment is elegant and might appear inviolate at the outset, a close examination reveals several problems with this thought experiment that diminish if not eliminate its capacity for demonstrating the supposed impossibility of strong AI. First and foremost, Searle makes a significant leap in logic when he determines that his own lack of understanding is analogous to the computer/program's lack of understanding. Searle essentially asserts that his feeding communications to the computer and receiving responses to slide back under the door is the same type of automatic and rote process that the computer itself would be performing, yet the two actions are enormously different in terms of complexity and in terms of product and effect. Searle's "non-understanding" action is purely mechanical, consisting of moving paper from where it is slipped under the door to the computer and then moving the output pieces of paper back to the door. Only part of computer and program's actions are mechanical, and then the least impressive and least "intelligent" parts -- those that move the paper, conduct the printing, etc. The processing of language requires a new task each and every time, and though the rules applied might be the same they would never be applied in the same pattern or lead to the same output (assuming different inputs). This action is thus fundamentally different from the action of a person feeding input to the computer and delivering the output, a task which never changes regardless of the communication taking place, and thus to suggest a lack of understanding in the latter is evidence of a lack of understanding in the former is entirely specious.

Second, Searle's argument can be broken down to a problem of semantics. It is not clear what "thinking" or "understanding" truly mean in Searle's context, and the outcome of the thought experiment is wholly if meaninglessly dependent upon these definitions. It can be agreed that what the computer/program does is not exactly the same as the human use of language in conversation, which is colored by emotional content, relationship between the speakers, larger external contexts, and more, but the computer/program in the Chinese Room is not "unthinking" simply because it has learned rules for syntax, vocabulary, grammar, and other mechanical elements of language. All human beings must also learn and utilize these language mechanics, and though humans do not receive a complete and comprehensive "rule book" of algorithms comparable to the program fed to the computer humans must receive instruction in some manner, even if it occurs unconsciously. This does not even disallow the theories of innate grammar forwarded by some, as even if some fundamental linguistic principles are hard-wired into the brain the specific rules and vocabularies for any given language can only be acquired through learning. That is, it is possible that language is itself an innate set of rules and mechanisms and thus that humans have a special capacity for language that could never be entirely emulated in an artificial machine/computer/program, and yet human language is still just as dependent on instruction and vocabulary building as is the program Searle imagines. The brain must take input and process responses (i.e. output) based on previous instruction in manner similar if not exactly analogous to that of the computer program in the thought experiment, and thus there cannot be said to be a "lack of understanding" simply because responses are generated based on previously learned rules of grammar, syntax and vocabulary.