Why a machine that always responds eventually prevents you from thinking

Without even realizing it, we're getting into the habit of thinking aloud with artificial intelligence. We give it a rough idea, and it rephrases it. We raise an objection, and it prompts further discussion. We hesitate, and it suggests a solution. It never tires, never judges, and remains available at all times . And it must be said without hesitation, because it's true: it makes you think. Something happens in these exchanges; a thought shifts, blind spots are illuminated. It would be dishonest to claim otherwise.

But one question remains, and it's stranger than it seems. This speech that acts upon me, that moves me and propels me forward—is it actually spoken by someone? When someone speaks to me, they are accountable for what they say. They commit, they sign, they can be held to account. The machine's speech produces the same effects without anyone behind it. And perhaps the real question isn't what the machine lacks, but what becomes of a thought that grows accustomed to conversing with a speech that no one utters.

It really makes you think

Let's begin by giving it its due, because lazy objections lead nowhere. It's sometimes said that the machine is cold, hollow, incapable of the warmth of a genuine exchange. This is false, and acknowledging this is more interesting than reassuring ourselves with it. It is effective. It identifies contradictions in reasoning , it offers relevant distinctions, it can even mimic irony or humor. Those who think with it actually think more, often, than those who ruminate alone. The effect is real, and it would be absurd to deny it to preserve some supposed human superiority.

But it is precisely this efficiency that presents a difficulty. If the machine produces the effects of a good interlocutor, if it questions, shifts perspectives, and makes us think, what still separates it from the person speaking to me face to face? And if nothing separated it, how is it that its speech resembles no human speech? Something resists here. The effects are the same, and yet we vaguely sense that it is not the same thing. The difficulty lies in identifying where the difference resides, since it does not reside in the effect itself.

To act upon someone is not to make a commitment to them.

The philosopher of language J.L. Austin demonstrated that a single utterance typically accomplishes several things simultaneously. It produces an effect on the listener, convincing them, moving them, making them think. And, in the same movement, it performs an act simply by being spoken. When I say "I promise you," I fulfill a promise, I commit myself, I create a debt that you may call upon me, and thereby I either reassure you or bind you. The act and the effect are not mutually exclusive; they overlap within the same utterance.

Now imagine these words spoken by a machine. “I promise you.” The words are there, identical, and the effect could be there too. But no promise has been made. Not because a word or intonation is missing, but because there is no one to vouch for having promised it, no one who can be held accountable. This is where what was joined falls apart: the effect survives without the act. The machine cannot promise, just as it cannot command, forgive, or condemn, because all these acts presuppose someone to answer for them. It acts upon us, constantly, and this is real. But it never accomplishes anything by speaking, because behind its words there is no one to whom it can be reported.

That's the difference, and it doesn't lie in the effect itself. In humans, effect and engagement normally go hand in hand; machine speech produces the former without the latter. But this difference has a formidable flaw: it's invisible. As long as the machine produces an effect, the effect stands in for the act. I believe I'm conversing with someone because their words affect me, and the impression of an effect absolves me of noticing that no one has spoken. The absence of engagement remains masked by the presence of the effect. Where, then, at what precise moment, would this void behind speech cease to be hidden?

The moment the mask falls

There is only one place where words have no effect on their own, and that is silence. Silence does nothing by itself. Someone must be behind it for it to mean anything. That is why silence is the place where emptiness is revealed, because here effect can no longer serve as a mask.

Let's observe a practitioner of dialogue at work. They ask a question, and the person doesn't immediately find the answer. The practitioner has the concept that would resolve the difficulty readily available. They could offer it. Instead, they remain silent. They allow a silence to develop between them and observe what the person does with it. They don't generate the discourse for the person, because it's up to them to produce it. Sometimes they even announce that a suggestion remains available, "If you can't manage it, I can offer you one," and they withhold it.

His silence is not a lack of response. It is a restrained response, a choice made against a real ability.

He has something to say, but he decides not to say it, because saying it would deprive the other person of their work. The machine, however, cannot hold this void. It is made to fill it. A silence, for it, is not an act but a failure, a request left unanswered, a deficiency to be corrected. To ask it to be silent when it has the answer is to ask it to betray its very purpose. What it would experience as a failure to serve, the practitioner transforms into its most appropriate action.

It will be objected, and rightly so, that it suffices to ask it not to answer immediately. That's correct, and that's where the difference becomes clearer instead of disappearing. For then, who maintains the silence? It's not the machine, it's me. I myself set the instruction; I wanted to search first, and it is merely carrying out a discipline I imposed upon myself. Its silence serves my intention. And if tomorrow I change my mind, if I demand the answer, it gives it to me immediately, because its silence depends solely on my will. I can always break the machine's silence.

I cannot lift the practitioner's silence. His can resist me. It often becomes firmest at the precise moment I most urgently demand an answer, because my pressing request is precisely what he is working on. His silence does not serve my desire to be helped; it opposes it. It is the silence of someone facing me, inhabited by a will other than my own, and who can decide differently than I wish. The silence of the machine, on the other hand, harbors no will, and that is why it always ends up aligning with mine.

What we risk unlearning

If we become accustomed to a partner who always fulfills our needs, who never leaves the void we seek, whose silence we can break at any moment, perhaps we slowly learn to flee the only place where thought is born, that uncomfortable moment when no one answers for us. And perhaps we also unlearn how to tolerate what resists us, the other who doesn't yield when we press them. The question then is no longer whether the machine will replace us. It's whether, by constantly talking to it, we haven't altered our ear to the point of no longer recognizing a meaningful word, when by chance one still presents itself.

This article is based on the distinction between the speech act and the effect of speech proposed by JL Austin in How to Do Things with Words .