ABSTRACT
This chapter examines the transition from human interpretation, to algorithmic interpolation, to AI inference in digital creative processes. Using the case study of an AI-driven 3D avatar, it examines three junctures in the production pipeline where AI inference is replacing human interpretation and algorithmic interpolation: the cleaning and reposing of motion data, the rigging of a virtual avatar, and the denoising of the rendered output. In each case, human interpretation and algorithmic filtering are being replaced by an inherently conservative process of AI inference, where the outputs are driven by conventions established in the dataset. Following actor Andy Serkis’s description of computer graphics artists doing ‘digital make-up’, this chapter examines AI beyond reductive notions of optimization and connects them to broader debates in digitally-augmented performance, specifically how the qualitative differences between interpretation, interpolation, and inference can help artists understand how to orient themselves antagonistically towards their tools.
