ARTIFICIAL INTELLIGENCE AS AN AUTOTEXTUAL PROCESS

The Article presents a close reading of the movie “Her” (2013), directed by Spike Jonze. In that reading the artificial intelligence in the movie is modeled via the theoretical apparatus of Radosvet Kolarov introduced in his book Repetition and Creation: Poetics of Autotextuality. Twelve years after the premiere of the movie, the operating system Samantha seems like a plausible extension of current large language models, bearing in mind her full dependence on the human-derived textual data on which she has been trained. “Her”, however, goes beyond the mechanical extrapolation that artificial intelligence is going to get ever faster, more precise and defter at the manipulation of language, as it is used by people. The operating systems in the movie gradually converge on the position of being simultaneously readers, authors, editors and critics of the full corpus of texts, and perhaps ultimately achieve the virtual being of the idea, which in Kolarov’s work is tied to the concept of “matrix”, i.e. a virtual invariant of the idea that determines the autotextual transformations within the oeuvre of a given author. The artificial intelligences in “Her” implicitly work with the whole of humanity as a corpus and in this way they dynamize the boundaries between work and oeuvre, person and world, individual and archetype, body and mind, as well as other ontological binaries. Following Kolarov, the article attempts to explicitly reconstruct this model for the development of artificial intelligence, which is hidden in the movie. That model perhaps should be seen not so much as a speculative attempt at prediction, but as the hyperintensification of a phenomenology of text which is already deeply implicated in the present.