This session explores the aesthetic, historical, and phenomenological implications of latent space and vector-based computational architectures in AI-generated media. As text-to-image and text-to-video models become embedded in standard media production workflows, latent space is an increasingly crucial site for understanding how technical infrastructures shape artistic expression, aesthetic judgment, and epistemological frameworks. Bringing together perspectives from media theory, aesthetics, and cultural history, the session investigates how latent spaces mediate notions of similarity, medium-specificity, and multimodality.

Presentations by Johan Malmstedt (Umeå University/Harvard University), Sebastian Breu (Humboldt University of Berlin), Sebastian Rozenberg (Linköping University), and Maja Bak Herrie (Aarhus University), with the participation of Leonardo Impett (University of Cambridge), will map the aesthetic, conceptual, and historical stakes of computational image generation. Together, the presenters examine how machine learning processes both challenge and expand established understandings of media, offering critical insights into the technical dynamics of AI-driven visual culture.

• Read program here

The session takes place as part of a network meeting in the DFF-Explorative Network AIsthesis but is open to all interested participants.

• Participation does not require registration.

Organizer: Maja Bak Herrie, Lotte Philipsen, and AIIM.