LATENT IMAGES

In a society rapidly turning to artificial intelligence to envision the future, humanity increasingly acknowledges its flaws and limitations, specifically the systematisation of human biases in machine learning. This affects every aspect of society to different extents, resulting in an urgent need to find ways of communicating how this technology operates.

This research addresses the formation of machine “knowledge” and its manifestation as (time, place and culture-bound) visual artefacts through a process of image-generation experiments with a network of machine learning models turning text into visual outputs. The process exposes the nature of the correlation between language and images in this environment and then investigates the new correlations introduced by further machine and human assessments. Bias is eventually recognised not only as harmful, but as a structural component of such knowledge, by observing how this technology reflects back onto humanity what humanity has fed into it.

MA Thesis: Hochschule für Gestaltung und Kunst FHNW Basel
Research+Design: Claudia Colombo
Year: 2021
Award: Swiss Design Award 2022 in Design Research