Literature, a True AI (Artistic Intelligence)

Critics of artificial intelligence have chosen the wrong target. A written or spoken text is, by its very nature… artificial. Language belongs not to biology, or even to the individual, but to culture: it is constructed, collective, and artificial. Language does not belong to us. Yet in writing literature, we manage to do a certain violence to language, tearing it from its context, which then allows the creation of a new object, artificial in its turn. But literary artifice admits itself as such, by revealing the artifice of every utterance.

Literature surprises us through its poetry (from the Ancient Greek poieîn, “to make” or “to create”), its way of recombining words and rhythms to produce unexpected and sometimes unconscious meaning. Literature is therefore a knowledge of recombination, of juxtaposition, of linguistic and narrative artifice. It is a form of intelligence, because it makes possible a new understanding of the world that links disparate and unusual perceptions (inter + legere, “between” + “to gather,” “to bring together”). The intelligence of literature has nothing to do with what is mistakenly called artificial intelligence. LLMs (large language models) generate simulations of human language using algorithms trained on immense databases. An LLM merely predicts the next word according to probability, that is, by choosing a word or a combination of words that has already existed in some form. The AI of LLMs cannot surprise us; its algorithm does not allow it.

The boredom we feel reading texts written by LLMs is the best proof that they cannot compete with literature. On the contrary, this so-called AI would be the annihilation of true AI, the artistic intelligence that is literature. The AI of LLMs reinforces the standardization of language by leaning on political consensus. The pleasure of writing, as of reading, is the possibility of breaking with predictable, flattened language, in search of other intelligences, other horizons.