WP1 Research Webinar: NCA for Artistic Exploration
Welcome to a webinar organized by WP1 titled NCAs for Artistic Exploration with Etienne Guichard.
Abstract
Etienne’s Presentation will cover Neural Cellular Automata (NCAs), their high-level applications in the machine learning community, their capabilities, and their potential applications to art. NCAs are a class of machine-learning models that broadly mimic biological cells. Derived from the simpler, highly complex, and interesting Cellular Automata (have a look at Conway’s Game of Life), they allow for biological-like, dynamic, and emergent behavior to be trained using data. NCAs have shown a remarkable ability to model cellular growth, complex image generation processes, problem-solving, and even artistic endeavors while being small, relatively easy to train, and often requiring less data than larger, more expensive machine learning models. These qualities allow NCAs to be trained and deployed locally. Artistically, NCAs present an opportunity for artist-model co-exploration. Their small size and frugal data requirements allow artists to maintain full sovereignty over their data. Their lifelike, dynamic behavior allows for the exploration of motifs, patterns, and styles from the artist’s work that have a life of their own. In this talk, an NCA-derived model (ART-NCA) will be presented. Exploring how Images can be used to train life-like behavior, their interesting properties, and discussing how this can be used for artistic exploration. Additionally, a conversation will be started with interested parties on how to further develop this model for visual and musical forms.
Bio
Etienne Guichard is a MishMash PhD researcher at Østfold University of Applied Sciences.
Access
Zoom link sent on WP1 mailing list. If you want to join, please ask for access.