A series of four digital print images, generated by using generative artificial intelligence. As a starting point for this work an original image coming from an analog video signal was digitised and used as a seed to start a feedback process. In this feedback process an image generating A.I. model was used as an echo chamber to visualise its own internal processes, which are unknown to its user. Researchers of large language models describe this as the misalignment between the computer model and their users. These models and humans have a certain expectation of one another: humans expect to have a reliable outcome whereas large language models are designed to minimise performance regression or a harmful outcome. As the feedback progresses we can see how the iterations are slowly regressing from each other but still reverberate visual signs of their previous versions.
Misaligned Feedback makes visible how these models are not explicitly designed to work with the ambivalence nature of abstract imagery.
2022 | Digital print on paper | Dimensions 60” * 60”