Could objects have a life of their own? The things around me, the lamp of my desk, the knob of the door, which shines as the white of an eye, the chair I am sitting on? Could they somehow be aware of my presence, and feel and think? Could a machine think? Not my computer (though I will not say it is dumb for fear it hears me and bugs), but a bigger machine that some secret agency would build in the dark underground of a mysterious cybernetic institute.

The question of whether a machine can think is a well-defined philosophical problem with a history dating at least from the 17th century. The question of whether, or in what sense, objects may have a life of their own, certainly introduces an uncanny aspect to the problem. In any case, these are the kinds of questions, philosophical or uncanny, from which Wittgenstein intends to liberate us by showing that they are nonsensical. By understanding that these questions do not make sense, we should be able to stop thinking about them. We should no longer feel apprehension touching the doorknob: it won't respond. We should know it is beyond the domain of what may be alive or dead, aware or unaware.

"Is it possible for a machine to think?" […] The trouble which is expressed in this question is not really that we don't yet know a machine which could do the job. The question is not analogous to that which someone might have asked a hundred years ago: "Can a machine liquefy a gas?" The trouble is rather that the sentence, "A machine thinks (perceives, wishes)," seems somehow nonsensical.  (Blue Book, 47)

In 1913, Heike Kamerlingh Onnes succeeded in liquefying helium. It brought him a Nobel prize. Thus, a hundred year before Wittgenstein was writing, one could ask whether a machine could liquefy gas. It seemed to be possible but no one could build such a machine. Is the problem similar when we ask for a thinking machine? It is not, according to Wittgenstein:

"Could a machine think?" … "Can a machine have toothache?" You will certainly be inclined to say: "A machine can't have toothache." All I will do now is to draw your attention to the use which you have made of the word "can" and to ask you: "Did you mean to say that all our past experience has shown that a machine never had toothache?" The impossibility of which you speak is a logical one. (Blue Book, 16)

Why can't a machine have a toothache? We might first answer that a machine has no teeth. But it is not exactly true, of course. The factory in Fritz Lang's Metropolis has teeth. An automatic saw has teeth. We just don't see how it could feel pain in its teeth, we don't know what the situation would look like. According to Wittgenstein, it is the same with a machine that would think. The idea is that we learn to use the words “think,” or “feel,” or “perceive,” by applying them to humans. We know the difference, when someone feels, or doesn't feel anything. When my friend is awake, I pinch him with a needle, he screams. When he is deep asleep, or if he is in a coma, I pinch him with a needle, he does not move: he can't feel anything. Or we see a philosopher in a painting, her head resting on her hand, her brow contracted: she is absorbed in deep thinking. We know when to apply these words to humans. But we don't know how to apply them to things. However, there are verbs we can apply to things, or animals, as well as to humans. A fish swims. A machine runs. A robot even walks. An automaton plays the piano. Thus, by grammatical generalization, because we cannot bear the irregularity of our grammar, where some verbs apply to humans and others do not, we try to apply words like “think,” and “feel,” and “perceive” to objects, and machines. We ask whether a machine could think, or an object feel. This is just what Wittgenstein calls a grammatical mistake:

“But surely a machine cannot think!" Is that an empirical statement? We can only say of a human being and what is like one that it thinks. We also say it of dolls and no doubt of spirits too.” (Philosophical Investigations, I, §360)

Verbs, like “to think,” “to feel,” “to wish,” but also words like “thoughtful,” “deceitful,” may have a certain extension and apply to things that look, more or less, like a human being (a doll, for instance). Then we know how to extend the implicit criteria that determine when something is thoughtful. But that extension is limited. Asking whether a machine, or a fish, or a chair, may be thoughtful is purely formal. The question stems from a grammatical generalization, and it cannot have an answer by itself. We just do not know when, in what circumstances, we could answer “yes” or “no.” We don't know what a thinking machine, or a thinking chair, would like.

“The chair is thinking to itself" … WHERE? In one of its parts? Or outside its body; in the air around it? Or anywhere at all? But then what is the difference between this chair's saying something to itself and another one's doing so, next to it? … We want to know how the chair is supposed to be like a human being; whether, for instance, the head is at the top of the back and so on. (Philosophical Investigations, §361)

Thus, the philosophical problem, whether a machine may think, and the more uncanny idea, that an object, a chair or the doorknob that seems to look at me, may be aware of my presence and, in that case, could even be malevolent, all would come from our being lost in the city of language. We can't find again the way we came to these questions. If we would understand the way we have constructed these questions, we would see, according to Wittgenstein, that they are meaningless, and we would stop wondering.

Let us then follow the guide again, Wittgenstein. In the end, these questions come from our not being able to accept the irregularity of the city of language. We try to apply to things words that were made for humans. But we don't have the criteria that would enable us to make sense of a thinking machine, or distinguish between chairs that would be aware of our presence and those that would simply not care. These ideas are just castles in the air. They don't make sense.

Is that enough to reassure me? Could I thus reason with myself, so as to be able to take in my hand the doorknob without apprehension, once I have gotten into my head the notion that it feels my hand, and detests this contact as much as I do?

Of course, when I first got convinced that the doorknob was looking at me, there was something by which I recognized that this doorknob in particular was alive. Maybe it was the reflection of the light (like the white dot the painter puts on the eyes of his portrait), or a slight resistance in my hand, which other doorknobs do not have. In fact, in the short paragraph quoted above, Wittgenstein raises all the questions that would have to have answered someone believing that the chair is thinking to itself... It could be for real or in a story. If you believe that the chair on which you are sitting is alive, and thoughtful, you must have somehow answered these questions: what shows that this chair is thoughtful? In which part of itself is it mumbling? How does it communicate, or not communicate, with yourself? It is what an uncanny story about a thinking chair would have told us. However, once these points have been answered, the thoughtful chair, or the problem of whether an object may think, or feel, or wish, makes perfect sense, for we have the criteria that enable us to extend to this object, or this kind of object, the vocabulary of thought. We know, we can depict, when the object is thoughtful, or when it is asleep, or if it is dead.

The difficulty is the same with the image of a thinking machine. There are numerous stories about thinking machines. Take Hal, the computer, in S. Kubrick's 2001: A Space Odyssey; there are moments when it hesitates, the lights on the panel in front of the astronauts shiver, its voice slows down. We understand, we see that the machine is thinking to itself, while talking to humans. Such stories have given us the conditions in which we could apply the verb “to think” to a machine, even though we cannot apply it to any machine in the real world. They have extended the conditions in which we can apply such verbs. Thus, they have made sense of the problem of whether a machine could think.

We tell ourselves stories. These stories extend the domain of applications of our words. They raise new possibilities, philosophical problems or uncanny ideas. But once these possibilities have been opened up, there is no going back. We know a computer may think, or a doorknob observe us.