The discussion of AI-generated art in the digital artist community often generates more heat than light, and I want to attempt a more useful framing.

AI image generation tools are a new category of tool with genuine capabilities and genuine limitations. The capability: producing coherent visual compositions from text prompts with remarkable speed. The limitation: the tool cannot make the aesthetic judgments that distinguish meaningful art from technically proficient imagery.

For working artists: these tools change the cost structure of certain tasks — mood board generation, rough thumbnail exploration, texture generation. This changes workflows but doesn't eliminate the need for artists who understand what makes images work.

For the broader ethics question: the training data provenance issue is real and unresolved. I don't think 'it's just like how humans learn' is a sufficient answer, because human learning and machine learning on scraped data are not equivalent processes. I also don't think 'all AI art is theft' captures the nuance correctly. The ethical analysis needs to be more precise than either slogan.