AI and the Automated Imagination

Shane Denson (Stanford University)

This presentation works through several recent attempts to address questions about artificial intelligence and (visual) imagination in an attempt to better understand the impact of AI on aesthetic experience. Drawing on the work of Don Ihde and Bernard Stiegler, Galit Wellner has put forward the concept of “digital imagination,” arguing that digital media, including AI, institutes a new phase in human imagination, moving us from a perspectival model of imagination to a “layered” model that is “co-shaped” by the new technologies and distributed between them and their human users. Also drawing on Stiegler, Yuk Hui argues that imagination has always in fact been artificial, and that we therefore have to take seriously the question of AI’s synthetic imaging powers as a force that could massively transform our own. I agree with both Wellner’s and Hui’s insistence on the historicity of the imagination and its openness to technical transformation. However, each of these thinkers leaves open a further set of questions, which I hope to address with recourse to Sartre’s early work on the imagination. The latter fleshes out, while de-essentializing, a dimension of Kant’s “productive imagination,” which mediates between intuition and understanding by way of a “schematizing” operation. As I will argue, Sartre’s contribution helps us to see this operation as deeply aligned with what Alexander Galloway has theorized as the new “visual contract” arising from computational imaging processes, a post-photographic contract that exceeds perspectival POVs to present objects from all sides at once. With this schematic view of computational images, we will be able to see the images produced by generative AI as embodying, in an important sense, an external homologue of the imagination and thus fundamentally changing our imaginative relation to the world—and not only for the better: the computational codification of schematizing operations significantly enables and constrains imagination and anchors political stereotypes (including those of race, gender, and dis/ability) in the external world. Finally, using this framework to look at artworks made with machine-learning algorithms will help us to better understand the experience, and the stakes, of our encounters with AI.

Shane Denson is a Professor of Film and Media Studies in the Department of Art & Art History and, by Courtesy, of German Studies in the Division of Literatures, Cultures, and Languages and of Communication in Stanford’s Department of Communication. He is currently the Director of the PhD Program in Modern Thought and Literature, as well as Director of Graduate Studies in Art History.

His research and teaching interests span a variety of media and historical periods, including phenomenological and media-philosophical approaches to film, digital media, comics, games, and serialized popular forms. He is the author of Postnaturalism: Frankenstein, Film, and the Anthropotechnical Interface (Transcript-Verlag/Columbia University Press, 2014) and co-editor of several collections: Transnational Perspectives on Graphic Narratives (Bloomsbury, 2013), Digital Seriality (special issue of Eludamos: Journal for Computer Game Culture, 2014), and the open-access book Post-Cinema: Theorizing 21st-Century Film (REFRAME Books, 2016). His book Discorrelated Images was published in 2020 by Duke University Press. Discorrelated Images explores the transitional spacetime between cinema and post-cinema. His latest book, Post-Cinematic Bodies, was published in June 2023 by meson press. The book asks: How is human embodiment transformed in an age of algorithms? How do post-cinematic media technologies such as AI, VR, and robotics target and re-shape our bodies?

Resgitration and information.