The Uncertainty is now mainstream
or is it Unlimited Creativity at your Fingertips, November 19, 2022
Well, in the week since the International Publishers Congress in Jakarta the AI and copyright uncertainty question has come out from the tall grass of the Twitter feeds, blogs and substacks, and is moving into the glare of media attention. If you have time for only one article, read the Verge story with a headline that could be the motto of this newsletter: “The scary truth about AI copyright is nobody knows what will happen next.”
The sequence was roughly this:
Nov 10/11 - Deviant Art, an online community for illustrators announced their offering of DreamUp™, a specially fine-tuned instance of Stable Diffusion to members of the community. They did also announce that members of the community could opt out of having their works included in the “fine-tuning” dataset that would give DreamUp its particular flavor.
Nov 12 - After asking of questions, the Deviant Art community awoke to the fact that before any “fine-tuning”, Stable Diffusion had already copied more than 1,000,000 images from the Deviant Art website to train the model (along with much of the rest of the visual internet it must be said). This comment from artist Megan Rose Ruiz captures the sentiment that echoed around Twitter.
Nov 15: Writer James Vincent puts out a sharp story on The Verge with a headline that could be the mission statement of this newsletter: “The scary truth about AI copyright is nobody knows what will happen next”. It sets out both the output question and the input question. It cites Andy Baio and Andres Guadamuz as being “surprised there haven’t been more legal challenges yet. ‘Honestly, I am flabbergasted,’ says Guadamuz.”
Nov 17: Wired publishes a story from founder Kevin Kelly on generative art, written with considerable tech-bro brio: “Picture Limitless Creativity at Your Fingertips”. It captures well the excitement that non-artists have about these models, but it paints anyone criticizing or questioning the development of these models as being losers caught in a tech panic cycle at the wrong end of history. This looks particularly bad in light of the Verve story and DeviantArt artists go after him on Twitter.
Nov 18: Bloomberg Law covers the story with“‘Wild West’ of Generative AI Poses Novel Copyright Questions”. They lead with the output question, ie 1) can an AI-generated work be copyrighted, and 2) what is the similarity test that would allow a creator to claim a violation when they see a work that is similar to one of theirs. it quotes people like Stanford’s Lemley as saying that infringement could only be claimed if the output infringes. But it does eventually get to the input question, and quotes Alex Touma of Allen & Overy LLP, “US law has not really addressed training AI datasets and whether any of that may be permitted by virtue of the transformative tests and fair use…I don’t think a lot of this has been tested in the court system yet.”
Where will we go from here? You tell me! Comments are switched on in the Substack app.