Blooming from Noise
Digging into the roots of AI-generated images
In an era of “black box” generative AI tools, concerns about ethical uses of AI are not just valid critiques but often a necessary pedagogical stance.
When every tool feels like it has the potential to do irreparable harm, it can be difficult to imagine classroom activities that meaningfully address text-to-image generation, especially with regards to environmental sustainability and exploitative data training practices.
This workshop introduces and uses the Common-Canvas model, built on open-source, Creative Commons materials. Additionally, participants will co-construct a small scale, low energy custom model built on locally sourced images of flowers. During downtime, while the model is training, we will have open discussion aimed at demystifying the underlying mechanics of diffusion models and imagining its implications in the classroom. We will cover a range of techniques and tools that can be applied for image verification and image generation purposes.
Workshop lead
Oscar Keyes, Ph.D. (he/him/his) is the Multimedia Teaching & Learning Librarian at Virginia Commonwealth University in Richmond, VA. His research investigates the emergence of cameras in education and their parallels with emerging technologies today. His pedagogical approach considers how social issues can be critically scaffolded into technical instruction for various digital tools, such as image-editing software, game engines, and generative artificial intelligence. He has taught (new and old) media arts in various spaces, including higher education, K-12 schools, summer camps, community-based arts organizations, and detention centers. When he’s not busy teaching, Keyes still makes movies with his friends.
Fig. 01 — Blooming from Noise, Oscar Keyes