QuestionsCan a machine think?Can machines think? As relevant as whether submarines can swim
Can a machine think?An ill-posed question

Can machines think? As relevant as whether submarines can swim

Submarines don't swim (Dijkstra)

Edsger W. Dijkstra 1930 · 2002Asking whether a machine can think is about as useful as asking whether a submarine can swim: better to study what programs actually do.
See profile
What it claims

The question "can a machine think?" is ill-posed: it is about as relevant as "can a submarine swim?". It is better to study what programs do and how they are built.

Why

Once we know how a machine works, whether it 'thinks' changes no fact: it is merely a question of vocabulary.

The same idea through history
Ludwig Wittgenstein1953In the Investigations (§ 360), Wittgenstein invites us to see the word "think" as an instrument and notes that the question "can a machine think?" concerns the use of that word.
Alan Turing1950Turing judges the question "Can machines think?" too meaningless to deserve discussion, and proposes to replace it with the imitation game.
Edsger W. Dijkstra1984Dijkstra takes up the remark in one sentence, in a talk at Austin: the question is about as relevant as whether submarines can swim.
In short

"Can machines think? As relevant as whether submarines can swim" is an argument by Edsger W. Dijkstra (The threats to computing science (EWD898), 1984, 1984) on the question "Can a machine think?" (answer: an ill-posed question). Its claim: The question "can a machine think?" is ill-posed: it is about as relevant as "can a submarine swim?". It is better to study what programs do and how they are built. 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, 0 of them with a documented reply, and the original text behind each premise, checked word for word.

The threats to computing science (EWD898), 1984
Valid with an implicit premise

In 1984, in a talk in Austin, the Dutch computer scientist Edsger Dijkstra criticizes the spirit of his field in its early days. According to him, the fathers of the field were rather confused: von Neumann compared computers and brains, and Turing looked for criteria to decide whether machines can think, a question of which "we now know" that it is about as relevant as whether submarines can swim. He does not develop the idea. The argument below is its strongest reconstruction: the image of a submarine moving under water shows that the answer to a question about a word changes none of the facts. A caveat: Dijkstra is not a philosopher, and he does not say that machines think or do not think. He says the question distracts from useful work.

P1Whether a submarine "swims" is a question of no interest. We know how it moves, how fast, how deep and what it is for. Whatever the answer, none of these facts changes: we would only be choosing a word.
P2The same holds for a machine and the verb "think": what a program does (its calculations, results, errors) can be described and checked without settling the question of the word. Turing himself judges it "too meaningless to deserve discussion".
P3 · implicit premiseA question whose answer changes none of the facts is a question of vocabulary, and it is not relevant to scientific or technical work.
Therefore the question "can a machine think?" is ill-posed: it is about as relevant as "can a submarine swim?". It is better to study what programs do and how they are built.

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 question that changes no observable fact but remains relevant: for instance whether the machine experiences anything.
Valid with P3. This premise is open to dispute: a question may change nothing about what the machine does and yet concern something else, for instance what the machine feels or understands. Searle's and Chalmers's objections go in that direction. Note also that the submarine analogy is Dijkstra's: it assumes that "think", like "swim", is a word whose use is a matter of convention.
Analogy. A fish swims, a man swims, a boat "cuts" through the waves, a submarine glides. Where does the verb "swim" stop? Nobody gains anything by deciding. The submarine image is Dijkstra's; this development is ours.

Objections

2

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

John Searle 1980 · attacks the conclusion

The question of understanding is not a question of words. A computer following a program manipulates symbols without knowing what they mean: it has syntax, not semantics. Whether a program suffices for understanding is, for Searle, "the right question to ask", and his answer is no. Here the answer changes what a machine is.

no reply
see the argument
We have not yet found a direct reply by Dijkstra in the texts.
Where the debate stands (our summary)Dijkstra sets aside the question of the word; Searle holds that the question of understanding is real and arises even if one does not dispute the word "think". Functionalists, for their part, reject Searle's argument. No consensus: everything depends on what one means by "understand".
David Chalmers 1995 · attacks premise 3

Even if we set aside the words "think" and "understand", a question remains that is not verbal: is there something it is like to be this machine? Chalmers speaks of "the really hard problem", that of experience. That an answer changes no observable behavior does not show that the question is empty.

no reply
see the argument
We have not yet found a direct reply by Dijkstra in the texts.
Where the debate stands (our summary)Open point. Dijkstra did not speak of consciousness, and we found no reply from him to this kind of objection. Some hold the question of experience to be real and distinct from that of words; others think it will vanish with a better understanding of the brain.