Eighteenth International Conference on the Arts in Society

Abstract

Machine learning models for text-to-image and image-to-image generation are used across disciplines to help visualize data and datasets in impactful and novel ways. Approaching human-computer creative collaboration from the perspective of a filmmaker generating moving image content, this paper describes how visual sequences derived from text-to-image and image-to-image processes using the Stable Diffusion model can be employed to imagine those who are absent from the historical record. Utilizing the life of an enslaved man who lived in southwest Virginia in the late 1700’s as a case study, this article describes how animations can be created by AI using carefully curated prompts contributed by the artist in the human-computer ensemble. This text proposes using human computer creative collaboration as a tool in the documentary production process to create representations of historical absences, or of those whose likenesses were never captured on canvas, paper or film. The study also proposes and defines the term ‘neuralpoiesis’ to describe new aesthetic forms resulting from human-computer creative collaboration. It can be argued that artists working as part of a human-computer ensemble are uniquely positioned to analyze and comment on the humanistic value of AI, especially as it relates to artistic production. Issues of bias are also discussed in regard of the tendency of AI to reproduce rather than eliminate bias due to data scraping processes. The importance of artistic experimentation with machine learning models to identify how they can be utilized to counter the oppressive nature of AI bias is stressed and examples are given.