Green Diffusion
This work depicts the interplay of destruction and creation by superimposing the process in which objects are broken down by soil microorganisms and then become fertile ground for new objects onto the diffusion model,
Diffusion model is a type of AI-based image generation, realized through learning the diffusion process of image data (in which noise is added, effectively destroying the data). By juxtaposing actual footage of physical decomposition in soil with AI-generated imagery, this piece highlights a new aspect of image generation.
In this installation, viewers are first diffused into a “sea of noise” using the same process as image generation. Then, using a diffusion model trained on the footage of soil decomposition, new images are generated.
Although microbial decomposition may appear destructive, it actually nurtures new life and shapes, just as an AI diffusion process emerges from the seemingly chaotic sea of noise to reconstruct new visuals. Presenting these two processes in parallel invites a fresh perspective on how “destruction” and “creation” intersect in both the physical and virtual realms.



