(C) Daily Kos This story was originally published by Daily Kos and is unaltered. . . . . . . . . . . GenAI Cafe [1] ['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.', 'Backgroundurl Avatar_Large', 'Nickname', 'Joined', 'Created_At', 'Story Count', 'N_Stories', 'Comment Count', 'N_Comments', 'Popular Tags'] Date: 2023-03-18 One of the main arguments artists have against GenAI is their artwork was collected off the internet without their consent or compensation. This is open to interpretation and contains a kilometers-wide gray area which will hopefully be narrowed down during the court cases to come. For now lets discuss the relevant issues: Ownership and authorship: As it stands right now Stable Diffusion is possibly in the clear because it collected images off the internet to learn how to devolve a field of static then back into the original image. Stable Diffusion would take the Mona Lisa, turn it into colorful static, then convert that static back into the Mona Lisa. It isn’t a direct copy, but it was recognizable as the famous painting. If still alive da Vinci would retain ownership of his painting, and as long as nobody tried to claim the GenAI as the real thing you could print it out and hang it on your wall. Legal status: Legally speaking, you can not copyright a style, technique or method of creating art. Thousands of artists have studied Michelangelo down to his individual brush strokes and used that knowledge to create their own paintings without fear of lawsuits. Even though this is what Stable Diffusion has done there are many artists who claim the algorithm has appropriated their unique style illegally, but it is indeed legal. Ethical considerations: The ethics side of the matter depends largely on what you do with the Stable Diffusion after it learns how to make art. Artists have often copied the style of other artists to present their visions, for example it is difficult to find a modern rendition of Art Nouveau that isn’t heavily influenced/copied from Mucha or the others. I recall seeing a painting that was extremely close to Dali’s work but the artist used cats instead of clocks. Even though the result was practically a brush-stroke by brush-stoke copy, it was still an ethical use of someone else’s style. Fair use: The images gathered from the internet were freely available to view. They did not break into artist’s private cloud depositories or trespass on paid-access sites. The artists put their work up for anyone to look at and study, which is what Stable Diffusion did. However, there were some sites that specifically warned their images were not to be used in any way other than as a representation of their catalog, which is the reason behind the Getty Images lawsuit. Compensation and attribution for human creators: Do the artists demanding compensation pay other artists themselves? When they study other artists and incorporate what they learn into their own style do they give proper attribution? Usually, no, they don’t. In the above example of the Dali-style melting cats the artists did not credit Dali for the inspiration as he assumed everyone would know of the original painting . Furthermore, how are you supposed to attribute a mix of styles? I often try multiple styles in my prompts to see what results the AI comes up with. For example: Vermeer oil painting of demon kitten in a teacup in the style of Tex Avery and Peter Max. The final image looks nothing like the three artists individually, so is attribution needed? Copyright for AI: The US Copyright Office has already stated that GenAI creations on their own can not be copyrighted, but how much human involvement is required before it’s copyrighted? Or even accepted as art? If you have ChatGPT write a descriptive paragraph of a sunset then you paint that sunset strictly following the text, is that copyrightable art? If you compose a poem, feed it through an AI Art bot, then work on the prompts until you get it just as you want, is that enough human activity to qualify? If you train your own GenAI with your own personal style (as one notable artist is doing as he slowly loses his hand dexterity) then does that count as “your” artwork? Transparency: There is a call for details surrounding the creation of AI-generated artwork, including the original data used to train the algorithms and the parameters used to generate the artwork, however this could prove impractical. Stable Diffusion used millions of images to train its algorithms, is it supposed to list every artist/source/reference in the image metadata? How the algorithm learned should be transparent, as is how it’s used, but beyond that isn’t it up to the end user to decide how much to reveal? [END] --- [1] Url: https://www.dailykos.com/stories/2023/3/18/2159035/-GenAI-Cafe Published and (C) by Daily Kos Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified. via Magical.Fish Gopher News Feeds: gopher://magical.fish/1/feeds/news/dailykos/