An incredibly average face : approaching 31,200 images with human eyes, sorting systems, and collaging

Abstract

This essay presents a practice-based research project investigating how beauty is defined in the context of image-generating AI. Using an automated prompting system, the author generates 31,200 images of faces that vary across four parameters (beauty, object, style, and seed). In regard to the question, how human actors can look at such large sets of images she develops systems to sort, filter, analyze, and visualize the resulting images. These include a custom file browser, UMAP visualization (PixPlot), 3D matrix representations, and algorithmic collaging. Looking at 31,200 images ultimately leads her to questions about value and digital waste.