Jon Chambers

Approximations of a Body Part

Machine Learning Algorithm, Digital Video


5 minutes

Approximations of a Body Part is a machine learning (ML) experiment that takes ML model outputs and uses them in other ML inputs. An earlier ML piece called BodyGen uses a custom ML model trained on images of 3D scans of Chambers' body that generates unsettling chimera images of body parts that don't exist but are still in the visual style of the original source images. The outputs from this custom model are then used as an initial style reference for the ML image/text to video experiments in Approximations of a Body Part. These new ML outputs create another generation of disturbing body parts similar in style to the original pink outputs of BodyGen, but start to move away from a specific readable body. Images from these new videos are then used as style references for another generation of image/text to video model inputs where the outputs become an approximation of an approximation of an approximation of a body. After a few cycles using this process, the resulting images and videos depict terrifying, seductive and uncanny distant echos of an original source. They become physical and psychological bodies struggling to resolve themselves within algorithmic surveillance, extraction and analysis of machine learning networked systems.