Gallery Entry

Origin Movement Rupture Echo – Image, Text, and Machine Interpretation

By Marcel-Nikodem Dziewulski · AT

What happens when an AI is asked to interpret the same image again and again? Origin Movement Rupture Echo explores how images change when they enter a continuous loop of interpretation. The project began with archival photographs from the Google Data Center Photo Gallery, depicting the technical infrastructure behind much of our digital everyday life. Rather than treating these photographs as neutral documents, I developed a method for examining how their meanings shift through repeated interpretation. The method consists of five phases: Origin, Movement, Rupture, Echo, and Reframing. It was developed by translating an analytical structure from Didier Eribon’s Return to Reims into a visual process. The method asks where an image comes from, how it changes through movement and transformation, what returns, and what remains. I then automated the process using the AI model Gemini. The system describes an image in text, generates a new image from that description, describes the new image again, and repeats the process. Image becomes text, text becomes image, and the resulting image becomes the starting point for the next iteration. The aim was not to produce a better or more convincing image. The repetition made it possible to observe the system itself. Across thousands of iterations, certain elements became more dominant while others disappeared. Ambiguous details were simplified, unusual forms were replaced by familiar ones, and recurring visual types became increasingly stable. The process revealed preferences within the model’s interpretation: what it tends to recognise, what it leaves out, what it considers important, and which visual references it repeatedly returns to. The iterative process therefore becomes a way of looking at the AI through its own output. Instead of analysing the system from the outside, its patterns become visible through repetition. Cultural references, aesthetic conventions, and statistical tendencies that remain difficult to see in a single generated image become increasingly apparent across a longer sequence. This approach is informed by Bertolt Brecht’s reflections on industrial photography, particularly the idea that a photograph of an industrial site can show its surface while concealing the social relations and conditions behind it. The project extends this question to machine interpretation: what happens when an image is not only photographed, but repeatedly translated by a system trained on vast amounts of existing visual culture? The resulting images are not simply the work of the AI. The project depends on the decisions I made in developing the method, selecting the source material, structuring the iterations, determining what to observe, and deciding which transformations and patterns became relevant. The AI is therefore neither a neutral tool nor an autonomous author. It becomes a participant in a process whose conditions are deliberately constructed and critically observed. This also creates a tension. Making algorithmic patterns visible can expose how generative systems reproduce existing visual norms, but reproducing these patterns can also reinforce them. The work therefore moves between analysis and reproduction, between revealing visual conventions and producing them again. The project ultimately asks what happens to visual culture when creative processes increasingly rely on the same models, datasets, and statistical systems. If these systems keep favouring familiar images, could our visual world become more similar over time? Origin Movement Rupture Echo does not attempt to provide a definitive answer. Instead, it constructs a process through which these questions can be experienced and examined. The resulting work was documented across 904 pages and can also be experienced as an audiovisual installation, where the images unfold one after another in a sound-responsive projection.

Category

CONCEPTS MIXED MEDIA

Approved

6 October 2026

Portfolio

https://marcelnikodem.com/