QuestionsCan a machine think?We say "it thinks" only of a living being, or of what resembles one
Can a machine think?It takes a living body

We say "it thinks" only of a living being, or of what resembles one

Wittgenstein: can a machine think?

Ludwig Wittgenstein 1889 · 1951The limits of my language mean the limits of my world.
See profile
What it claims

For such a machine, saying that it thinks has no place in our use of the word. The question is not to look for a hidden thought in the machine: it is to ask how far it resembles, by its life and conduct, what we say thinks.

Why

We say "it thinks" only of a living being or of what resembles one; a machine falls outside how the word is used.

The same idea through history
Ludwig Wittgenstein1953The starting principle: the meaning of a word is its use in the language. So we must look at how "think" is used.Ludwig Wittgenstein1953Sections 281-284 are at the heart of the private language argument: "pain" is learned about what behaves like a human being, not about a hidden object.
In short

"We say "it thinks" only of a living being, or of what resembles one" is an argument by Ludwig Wittgenstein (Philosophical Investigations, §§ 281 and 359-360, 1953) on the question "Can a machine think?" (answer: it takes a living body). Its claim: For such a machine, saying that it thinks has no place in our use of the word. The question is not to look for a hidden thought in the machine: it is to ask how far it resembles, by its life and conduct, what we say thinks. According to AskPhilosophy's logic check (Z3 solver), the argument is valid only if one grants a premise the author leaves unstated (P3). AskPhilosophy presents 2 objections, 2 of them with a documented reply, and the original text behind each premise, checked word for word.

Philosophical Investigations, §§ 281 and 359-360
Valid with an implicit premise

When we ask whether a machine can think, we imagine a fact to discover: is there, inside the machine, thought, yes or no? Wittgenstein proposes a different look. In his Investigations (written about 1945-1949, published after his death), he notes that we say "he suffers", "he sees", "he thinks" of the living human being, and of what resembles one and behaves like one. Saying that a machine cannot think is therefore not stating a fact of experience: it is noting how the word "think" works. A caveat: Wittgenstein never concludes "so no machine will ever think". This argument is a reading of his remarks, defended notably by Wittgensteinian commentators, not a thesis he states in one sentence.

P1One can say only of a living human being, and of what resembles one and behaves like one, that it has sensations, sees, hears, is conscious (§ 281). The same goes for "think": "We say only of man, and of what resembles him, that he thinks" (§ 360).
P2"A machine cannot think" is not a proposition of experience, which one could check or refute by opening the machine. It concerns the use of the word: one must look at the word "think" "as an instrument" (§ 360).
P3 · implicit premiseA machine, as we know them (a manufactured device that calculates), is not a living being and does not behave like a living human being throughout its life.
Therefore, for such a machine, saying that it thinks has no place in our use of the word. The question is not to look for a hidden thought in the machine: it is to ask how far it resembles, by its life and conduct, what we say thinks.

click a premise to reject it

If you accept every premise, logic forces you to accept the conclusion. So the debate is about the premises.
Checked by a logic solver: the premises cannot all be true while the conclusion is false.
Without the implicit premise: Without P3, the computer finds a machine that resembles a living being enough for "it thinks" to make sense: the conclusion falls.
Valid with P3. This premise is the most fragile: what counts as "resembling"? If a machine one day behaved like a human throughout its life, P3 would fall and the conclusion with it. Wittgenstein does not close the door: in § 359 he notes that the human body "comes closest to being such a machine".
Analogy. The word "win" is used for games, trials, elections. Asking whether a pebble can "win" is not a physics question about the pebble: we see that there is no game in which the pebble plays. This image is ours, not Wittgenstein's.

Objections

2

Each objection is a conversation: click to see who replies and where the debate stands.

Alan Turing 1950 · attacks the conclusion

Turing judges the question "Can machines think?" too vague to deserve discussion, and replaces it with a test: can a machine pass for a human in conversation? He also predicts that by the end of the century the use of words will have changed so much that one will be able to speak of machines thinking without being contradicted. Usage is not fixed: it will follow what machines do.

a defender replies
see the argument
We have not yet found a direct reply by Wittgenstein in the texts.
S
Stuart Shanker defender of the argument · 1998
Summary: Shanker reads Wittgenstein's late remarks on psychology as a critique of the idea that thinking comes down to computing. A change of usage would not be the discovery that machines think: it would be a change of concept, and one would still have to say what is gained or lost in speaking that way.
Where the debate stands (our summary)Turing and Wittgenstein agree on one point: the question concerns the use of the word "think". They differ on what follows: for Turing, usage will extend to machines that pass the test; for Wittgenstein's heirs, extending it would be a choice of language that needs justifying, not the discovery of a fact. Open point.
Hilary Putnam 1967 · attacks premise 3

Putnam defends functionalism: a mental state (such as pain) is defined by its role in the organization of the system, not by what it is made of. The same pain could be realized in a human brain, in an octopus or in an electronic system. Being alive is therefore not required to think: only the right organization matters.

a defender replies
see the argument
We have not yet found a direct reply by Wittgenstein in the texts.
Hilary Putnam defender of the argument · 1988
Summary: Putnam himself gives up functionalism in Representation and Reality. His argument: any physical system can be described as realizing any abstract organization. The expression "same organization" therefore does not suffice to fix what thinks and what does not.
David Chalmers 1996
Summary: Chalmers replies that Putnam describes too easy a realization. For a system to realize an organization, its states must be linked by causal relations that also hold counterfactually (what would happen with other inputs). With this constraint, a rock does not realize just any automaton.
Where the debate stands (our summary)The basic disagreement is about what makes a system fit to think: its organization (functionalism), or also its life and history (the living position). Functionalists have fended off the too-easy-realization argument; whether organization suffices remains open.