If you have never looked at AI from that angle, you might wonder why anyone would question the ethics of artificially generated images at all. Text-to-image generators are a lot of fun. You enter any string of words and they come up with a visual creation that corresponds to your input. All that within a couple of seconds.
But have you ever thought about how they are actually doing that? Image synthesis models (ISMs) like Midjourney, DALL-E, Stable Diffusion and Google Imagen learn to generate images by analyzing millions of images they find online.
Their databases have been built using photos, drawings, sketches, technical drawings – and artistic styles and artworks. Anyone’s artworks. Including those of the most famous, long-dead artists in history. Based on those data sources, artificial intelligence generates images that it associates with the user’s text input.
Generative AI never replicates any particular artwork; the algorithms are programed to not do so for copyright reasons. But the sourcing process alone raises major ethical and privacy concerns because the images are taken off the web without permission and proper attribution to the authors.
Another qualm about AI art is that it so easy to create images that mimic specific art movements. Or, the style of a particular artist who might have spent years developing their unique art style.