Artificial Intelligence and Image Politics: Memory, Erasures, and Dissidences

Abstract

This paper analyzes the radical transformation of the visual regime brought about by generative artificial intelligence, characterizing it as a shift from photographic representation to statistical simulation. It argues that AI systems, by training on biased datasets, perpetuate a "machinic eugenics of the gaze" that reinforces racism, ageism, and gender norms. The concept of "latent space" is explored not merely as a technical domain, but as a site of tension between algorithmic erasure and the potential for dissident visualities. Through the discussion of the author's artistic research projects (specifically Botannica Tirannica and Poisonous, noxious and suspicious) and the case studies of Maria Bandeira and Luzia Pinta, the text demonstrates how AI can be used to fabulate archives for historically silenced narratives. Finally, it proposes a perspective rooted in the Global South that views technology as a tool for critical imagination and the activation of potential histories.