Why Doesn't AI Respond the Way I Want?

In Stalker, the masterpiece by Andrei Tarkovsky, the Zone is a mysterious place where, according to legend, there is a room capable of fulfilling the deepest wishes of those who visit it. One of the travelers, the Professor, wants to verify the real existence of the room. The other, the Writer, doubts his purpose, fearing what he may discover. But it is the Stalker, the guide, who understands the spiritual dimension of the journey: it is not simply about a conscious desire, but about the truest wish, even the one we do not know we have.

An emblematic case told in the film is that of Porcupine, a stalker who took his dead brother to the Zone hoping to bring him back to life. When he leaves, he receives an unexpected fortune: the room granted his most intimate desire, not the one he thought he had. Overwhelmed by this truth—his real wish was money, not his brother—he commits suicide. He could not bear the distance between what he said he wanted and what he truly desired.

This abyss between the question asked and the hidden wish is repeated today in the relationship between humans and artificial intelligence. Someone who asks “what is the state of global geopolitics?” may actually be looking for strategies to protect or multiply their assets. Someone seeking an explanation of Nietzsche may be trying to understand their own existential anxiety. The question is rational, public, acceptable. The desire, on the other hand, is intimate, confused, sometimes unsharable.

AI, like the room in Stalker, responds to what is asked of it. But it cannot always, nor should it, guess the deep desire that motivates the question. Yet, a revealing tension exists: the quality of our questions is determined by the clarity—or lack thereof—regarding our true aims. Like Porcupine, we often do not know what we really want, and that contaminates our searches.

In a world of intelligent assistants, the key is not to ask better questions, but to wish better. Understanding what drives us, what we deeply long for, can redefine not only how we interact with technology, but also how we relate to ourselves. Otherwise, we might end up like Porcupine: discovering too late that what we asked for was not what we wanted, and that the right answer may, in fact, be a sentence.