Alan Turing, the Imitation Game and What Got Lost in the Swap
“Changing the question is not the same as answering it, though it can sometimes look remarkably similar.”
In October 1950, the philosophy journal Mind published a fourteen-page article by a British mathematician who had helped crack the German codes during the war and who, a few years earlier, had laid the theoretical foundations of modern computing. The article was titled Computing Machinery and Intelligence, and its opening sentence proposed considering the question: can machines think? What Alan Turing did next, over the following pages, is less well known than it ought to be, and it is precisely what deserves examination here: he did not answer that question. He replaced it with another.
The Imitation Game, as Turing Described It
Before turning to machines, Turing describes a parlour game: a man and a woman each enter a separate room, and an interrogator in a third room must work out, by means of written questions, which of the two is the woman. The man tries to mislead the interrogator; the woman tries to help him identify her correctly. Turing then proposes a variation: what would happen if a machine took the man’s place? If the interrogator is wrong about the machine as often as he was wrong about the man, Turing says, then it makes sense to say the machine plays the game well. Only after setting this out does he arrive at the sentence that would shape the course of artificial intelligence for decades to come: that this new question — whether a machine can play the imitation game satisfactorily — usefully replaces the original question, which he himself considered too vague to deserve discussion, of whether machines can think.
A Substitution, Not an Answer
It’s worth pausing on that move, because it is exactly the same one this series has been pointing to since its first instalment. Turing does not argue that indistinguishable behaviour proves the presence of thought; he argues that the question of thought is so confused that it is better abandoned altogether, leaving only behaviour, which can actually be measured. It is an honest manoeuvre, and Turing admits as much without evasion: he changes the subject because he finds the original topic unresolvable. The trouble is that much of the culture that inherited his article failed to remember that honesty. It kept the substitution and treated the outcome of the imitation game as though it were, in fact, an answer to the question Turing had chosen to set aside.
The Objections Turing Himself Had Already Anticipated
What’s striking is that Turing himself, in that very article, pre-empts several objections, including two that matter to this series. To what he calls the argument from consciousness — that no one can know a machine thinks unless they feel, as they do themselves, that they are thinking — he replies that such a standard, applied consistently, would lead to solipsism: the idea that I can only be certain my own mind exists, since we cannot verify another person’s consciousness either, and yet we adopt, as a practical convention, the belief that others do think. To what he calls the theological objection — that thinking is a function of the immortal soul, and God gave a soul only to man, not to animals or machines — he responds that limiting God’s power in that way would be, in his words, an impious restriction: if God can give a soul to an elephant, then in principle he could also give one to a machine, and no one is in a position to decide on his behalf where that possibility stops.
Why That Theological Reply Falls Short
This last objection deserves a better answer than a simple dismissal, because Turing is right about something: no one can limit in advance what God would be capable of doing. But that is not the point the Christian tradition is making when it speaks of the image of God, and conflating the two is Turing’s error here. The question is not what God could do in theory, but what he did do, according to what he himself chose to reveal. Genesis 1:27 does not describe a general, transferable capacity that God might grant to anything sufficiently complex; it describes a particular act, narrated, directed at a specific creature at a specific moment in history. The doctrine of the image of God does not rest on its being metaphysically impossible for God to do otherwise with some other creature; it rests on what Scripture affirms he actually did. Turing answers an argument about the limits of divine power, but the real Christian objection is not about power at all — it is about revelation. It is not that God cannot; it is that we have no indication he has done so with anything other than human beings, and all the biblical evidence points the other way.
Conclusion
The foundational error was not that Turing proposed the imitation game, which remains a legitimate and valuable philosophical exercise. The error was that several generations after him treated that game as though it were equivalent to settling the question Turing had, quite honestly, chosen to set aside. Decades on, we are still asking ourselves whether a fluent conversation is enough to recognise someone on the other end of the line, and we still, all too often, make the same slip: mistaking having changed the question for having answered it.
