The Dividual Enunciation. Rethinking the Production of Images in the Age of Transformers

The concept of enunciation has long held – and continues to hold – a central role in semiotic studies, serving both as a theoretical framework and as a tool for analyzing texts, discourses, and signification processes. This paper seeks to apply the concept of enunciation to Generative Artificial Intelligence (GenAI), particularly Visual Generative Artificial Intelligence (VGenAI). My intent is twofold: on one hand, I intend to develop a deeper understanding of how algorithmic machines operate in the automated creation of images; on the other, I want to test the flexibility of the theoretical notion of enunciation and to identify the transformations that VGenAI introduces into this concept. My central hypothesis is as follows: traditionally, reflections on enunciation have emphasized the mutual individuation between utterances and their subjects, a kind of identification enacted through discursive production practices. In contrast, GenAI and VGenAI offer a different model based on operations of dividuation – that is, the arbitrary partitioning of both discourse and its subjects. This new scenario, on one hand, connects enunciative practices to a broader economic and political context; on the other, it prompts a reconsideration of the fundamental concepts of enunciation theory.