In this Screen Walk, Leon-Etienne Kühr and Ting-Chun Liu will guide us through their experiments with feedback-driven generative AI images in a live coding session.
Machine learning systems rely on statistical models that learn from examples and estimate how to reproduce them. The interface for generative AI increasingly operates on a simple input-output principle, reinforcing the illusion of intelligence. However, closer inspection reveals that these are complicated pipelines of interconnected models and algorithms, each influencing the data and steers the outputs in unique ways.
Using feedback as a strategy and cybernetics as a framework, Kühr and Liu deconstruct these pipelines. Through programming and analysis, they will show how individual components drive AI generated images and reveal the hidden mechanisms that guide their visual outputs.
Leon-Etienne Kühr is a computer scientist and media artist leading the AI Lab at the Offenbach University of Art and Design. His work spans information visualization and data science, focusing on the exploration and visualization of datasets and models that are increasingly embedded in everyday life through AI-driven automation. His artistic research examines how our evolving algorithmic landscape might shape a future where human-machine interaction gives way to machine-to-machine feedback loops.
Ting-Chun Liu is a media artist, researcher, and self-described half-baked programmer from Taiwan. He currently serves as the artistic associate at the Bauhaus-Universität Weimar. His practice revolves around audiovisual media, network practices, and artificial intelligence. His research focuses on algorithmic art and critical AI, exploring the cybernetic mechanisms in generative systems.
Screen Walks is a series of live-streamed artist/researcher-led explorations of online spaces and artistic strategies designed to illuminate a thriving – often overlooked – digital cultural scene. A new online collaboration between The Photographers’ Gallery, UK and Fotomuseum Winterthur, Switzerland.