The relationship between text and image has been a longstanding concern of aesthetic theory. On the one hand, it has invited strong formalist calls for their separation as entirely distinct media whose aesthetic integrity requires that they be kept separate; on the other hand, such dreams of purity have been rebuked as misguided because “there are no visual media” (W.J.T. Mitchell). This talk proposes that state-of-the-art instances of “multimodal AI” both complicate the relationship between text and image and help us to rethink it. Multimodal AI – particularly text-to-image machine learning systems like Dall-E 2 or Stable Diffusion – collapses the distinction between text and image by circumventing this separation altogether; at the same time, it brings both mediums into a new performative constellation: language here produces images in a process of “operative ekphrasis” that is distinct from older, strictly algorithmic techniques. The talk develops this concept by way of examples from visual poetry; argues for a new role of language in our digital present; and shows how new technologies both require conceptual analysis and can themselves serve as catalysts for conceptual innovation.
Hannes Bajohr is currently a Junior Fellow at Zurich’s Collegium Helveticum. His research spans German intellectual history in the 20th century, experimental literature after cybernetics, and theories of the digital with a focus on Critical AI Studies. In 2017, he received his doctorate from Columbia University with a thesis on Hans Blumenberg’s theory of language. He co-edited the Hans Blumenberg Reader (Cornell UP, 2020) and the correspondence between Blumenberg and Hans Jonas (Suhrkamp, 2022). His most recent book is Schreibenlassen (August Verlag, 2022), a volume on digital literature, post-conceptual practices, and the critique of aesthetic AI.