The term “unhistory” refers to “stories of ordinary people who are not considered historical.” Given that AI systems work by compressing information into patterns through the removal of data points that do not fit the main pattern (or Ordinary), this mixed-media work sees unhistory as a provocation – “who or what is ordinary?”
Exploring this, the artist has assembled an autopoietic fictioning system, based on a dataset of 80 second-hand archival 35mm slides. Before home video cameras were widely used, 35mm slides were a way for people to share memories and experiences with the image acting as the “prompt” for a nostalgic recollection of the past.
The slides in this work were purchased from anonymous sellers online, detritus from family archives and personal collections. The artist requests the AI system observe the slides through a camera. The system identifies visual features and weaves these into a fiction, what it considers an “ordinary” (in the statistical sense) memory based on these scenes. The memory is recounted to the listener through a generative voice, which is combined into an evolving soundtrack that is listened to via headphones.
The slide projector system works in real-time – if a viewer interacts with the projected image, their silhouette will be observed, and change the narrative created, creating a new “unhistorical account”.
The tension between the machine responses and our own perception of these same artefacts begins a process of “unlearning” (Chun) – a diffractive method revealing patterns of difference within dominant historical narratives and potential for their transformation.
The machine learning models that generate these stories are amalgams of masses of human-data, and used widely in predicting and shaping the future. As such they are hegemonic instruments of power, and one way of revealing the distortions and oddities in the predictive “power” of these systems is through prompting them to tell us stories about what we perceive directly.
Accompanying the AI system are a series of dual-layer acetate prints, based on reproductions of slides in the dataset. The second layer of the print is a custom process that keeps the ink fluid and allows it to flow widely before drying. The fluid dynamics, surface tension and agglomeration enact a material instantiation of the data sorting processes that are inherent within AI systems. The “ground truth” datapoint printed on the first layer contrast with the “compressed” fluid second layer work, enabling a diffractive reading of the scene depicted.
Exhibitions
Goldsmiths Computational Arts MFA Show September 2023
