The Latent Path

The Latent Path is a 10-15 minute real-time performance that uses improvised AI-generated images to complement the live reading of a text. The text–written and read aloud by my collaborator Emrys Brandt–is a poetic meditation on the act of meaning-making through the production of maps, and the navigation of physical / virtual / latent space. A trained machine learning model, like the text-to-image AI model Stable Diffusion used in this performance, consists of a ‘latent space’, which can be thought of as a high-dimensional space that represents all possible images a model can produce. This work explores the notion that generating AI images can be thought of metaphorically as a process of navigation–of wandering through a space of possible images as one might wander through a city.