First session: Facing the Mediasphere
31 mrt. 2023
Abe Geil, University of Amsterdam
This Person Does Not Exist (TPDNE) is an online project aimed at sounding an alarm about
the capacities of AI image generation. With each visit to the project’s webpage a new photorealistic portrait, seemingly indistinguishable from photos of actual persons, is produced by advanced deep learning algorithms called Generative Adversarial Networks (GANs). Announcing the presence of “this person” whose existence is simultaneously denied, the project enacts a variation of the epistemological crisis – the inability to distinguish between real and fake, true and false – through which such developments are
usually framed. This paper shifts away from that forensic framing by comparing the faces of TPDNE not to portraits of actually existing persons (nor to deep fakes) but rather to portraits of an older but equally non-existent person, namely the “average type” – that chimera of mid-nineteenth-century statistics that the Victorian eugenicist Francis Galton sought to visualize with his technique of composite photography. While a number of scholars have traced a line from Galton to contemporary developments in AI facial technology (Gates; Chun; Dondero), I seek to complicate that lineage by positing a cybernetic break between these two moments. Whereas Galton could only fantasize that his composites would fuse the empirical plentitude of photography with the abstract precision of statistics, with AI photography the image is arguably now fully subsumable within an operational logic of statistical abstraction. As a consequence, the modern binary of “individual” and “type” has undergone a chiasmatic transformation which effectively short-circuits its grounding in the logical opposition of concrete/abstract. While Galton sought to produce types as “real generalizations,” TPDNE generates images of singular individuals as when we might now call “real abstractions” (Toscano). In the process, the violence of abstraction at the site of personhood is relocated from the biopolitics of racialized types to the flexible, scalable, and automated production of the visual markers of identity for digital capitalism. In this conjuncture, I argue, the abstraction of personhood cannot be countered by simply insisting upon the concrete singularity of the individual. Rather, I suggest a rethinking of the category of type unhinged from its binary determination.