Note — Nov 06, 2022

Invasive Diffusion: How One Unwilling Illustrator Found Herself Turned Into an AI Model

Synthetic media (generative AI?) keeps making incredible leaps in the quality of renders and ease of use of its tools. Andy Baio tried Stable Diffusion with DreamBooth to find out how easy it is and spoke to both a copied artists and the coder who created the model copying her, to get their thinking around this most recent generation of tools. Baio also trains DreamBooth on 30 images of himself and gets frankly astonishing results representing him in various styles.

Reading the Reddit thread, his stance on the ethics seemed to border on fatalism: the technology is inevitable, everyone using it is equally culpable, and any moral line is completely arbitrary. […]

Ultimately, he thinks many of the objections to it are a misunderstanding of how it works: it’s not a form of collage, it’s creating new images and clearly transformative, more like “trying to recall a vivid memory from your past.”

More → Matt Haughey tried the same combination of tools and provides more technical details and a lot more stylistic variations. Make sure to check out his second batch of experiments, incredible.

+ Randomly, a few minutes later on things magazine I bumped into this, which I quite love: “AI is the real grey goo; what will be overwhelmed first is not physical space but cultural space”

+ Finally, a couple more things on generative AI (haven’t read those yet). Why ‘generative AI’ is suddenly on everyone’s lips and Sequoia’s Sonya Huang: Generative AI hype “absolutely justified”.