Hallucinations techno-imaginaires : une critique flussérienne des systèmes d'IA générative / Tecnoimaginando alucinações : uma critica flusseriana aos sistemas de IA generativa
The article examines so-called hallucinations in generative artificial intelligence models through the lens of Vilém Flusser’s philosophy of technology, interpreting them not merely as technical failures but as revealing symptoms of the structure and normativity of these systems. Drawing on Flusser’s notion of the apparatus as a black box, the article establishes a parallel between the photographic apparatus and AI models, whose internal functioning remains opaque to both users and developers. The text critiques the anthropomorphic use of the term “hallucination,” arguing that it euphemizes systemic errors and obscures their epistemological presuppositions. In contrast, hallucinations are understood as productive failures capable of exposing biases, regimes of truth, and economies of meaning within generative models. Mobilizing the concept of techno-imagination, the article analyzes artistic and poetic practices that deliberately explore these deviations as ways of playing against the apparatus’s program. It concludes that such practices establish zones of friction in which provisional modes of critical freedom vis-à-vis contemporary apparatuses can still be rehearsed.