In this talk, we present our collaborative investigations into the hidden mechanisms of generative image systems. Beginning with a paranoid methodology: while these systems often appear as black boxes with clearly defined inputs and outputs, peeking inside reveals intricate pipelines of carefully interwoven models and algorithms that produce legible images only through their complex interactions.
By obsessively dismantling the seemingly magical processes behind popular diffusion models, we reveal the constructed nature of the resulting images. We will discuss our practice of dissecting generative systems, investigating individual models, and applying various feedback and adversarial tactics at different stages of the image-making process to better understand the underlying mechanisms while finding aesthetic potential within them.
Leon-Etienne Kühr is a media artist, computer scientist, and research associate at the University of Art and Design Offenbach, where he leads the AI Lab. In his artistic research practice, he works with methods from information visualization, data science, and artificial intelligence. His work focuses on the mechanisms of generative AI systems and on the phenomena emerging from an increasingly entangled relationship between the world and its digital representations in generative systems. He studied media arts at the Academy of Media Arts Cologne and media informatics at Bauhaus University Weimar.
Ting-Chun Liu (Taipei, Taiwan) is an artist and programmer, and currently works as an artistic associate and lecturer at the Bauhaus University Weimar. His practice and research operate at the intersections of critical artificial intelligence, audiovisual media, and network practices. Through dissecting technological systems with feedback-driven interventions, he questions the media specificity of generative images, examines how perception, aesthetics, and power dynamics propagate through computational processes. He studied at the Academy of Media Arts Cologne and Taipei National University of the Art.
The talk will be held in English
Details
Date 16. January 2026
Time 5:30 pm - 7:00 pm
Finished


