The Chinese Room
On the axis: Can a machine think · a mind that understands
Locked in a room, one answers flawlessly in Chinese by following a rulebook, without understanding a word: is running a program ever understanding, or only giving every sign of it?
This problem is brought out by the thought experiment The Chinese Room:
You are locked in a room. Through a slot, sheets covered in Chinese symbols are passed to you, symbols you do not understand in the slightest. But you have a thick rulebook, written in your own language, that tells you: "to this mark, reply with that one." You follow the rules, you pass sheets back, and outside a Chinese reader finds your answers flawless: to them, the room understands Chinese. You, though, have understood nothing. A machine that handles our words, a chatbot that answers so well, is it doing anything other than what you do in that room?
Searle, Minds, Brains, and Programs (1980)
Is running a program, however flawless, enough to understand, or does handling symbols by their form alone stay blind to their meaning?
Your choice above is recorded in your profile. See my profile →
Justification
The : shut in a room, Searle manipulates Chinese symbols following a rulebook, and his answers pass for a speaker's, without his understanding a word. A program is only a syntax, a set of rules bearing on shapes; understanding further requires a semantics, a relation to the world that formal manipulation never supplies. Yet he does not conclude that only the living think: the brain is a machine, and an artefact reproducing its causal powers would think. What he refuses is therefore not the thinking machine, but the idea that running the right program suffices.
"Minds, Brains, and Programs" (1980)
Justification
Turing judges the question "can a machine think?" too dependent on the use of words to be settled, and replaces it with a test, the : an interrogator converses blind with a human and a machine; if they can no longer tell them apart, there is no longer a reason to refuse to one what is granted to the other. To the objection that the machine would still lack consciousness of what it says, he answers that pressing it leads to solipsism: I know others think only by their behaviour, and refusing that same evidence to machines alone is arbitrary. The substrate therefore matters little, a universal machine being able to simulate any discrete-state machine.
The original question, "Can machines think?", I believe to be too meaningless to deserve discussion."Computing Machinery and Intelligence", Mind (1950)
"Computing Machinery and Intelligence", Mind (1950)