Creative Assistants with Style: Making Sense of Generative AI as “Style Engines”

Abstract

Generative AI technologies have been heralded for their ability to become powerful assistants, creating plausible text and realistic images. Yet they have also frequently been criticized for their lack of precision, accuracy or veracity. We argue that focusing on such traits misses what is most novel and defining about generative AI. As probabilistic technologies, generative AIs do not store, in any traditional sense, any data or content. Rather, essential features of training data become encoded in deep neural networks as patterns, or what we refer to, as styles. We discuss what happens when the distinction between objects, their properties, and appearance dissolves and all aspects of images and text become understood as styles, accessible for exploration and creative combination. We suggest that the ability to explore the world with styles is a defining feature of generative AI, with significant implications for how we assess its usefulness as creative assistants.