Maybe something a little Baroque-ish?
Image: Battle of Actium by Laureys a Castro, 1672
Image: Christ in the Storm on the Lake of Galilee, by Rembrandt, 1633
Image: Dungeons & Dragons: Ghosts of Saltmarsh, by Greg Rutkowski, 2019
Interesting announcement today of a class-action lawsuit against Stability AI, DeviantArt, and Midjourney for their use of Stable Diffusion. The argument is that these AI systems have been (and are being) trained on copyrighted work. Further, the announcement says “…because all the visual information in the system is derived from the copyrighted training images, the images produced —regardless of outward appearance — are necessarily works derived from those training images.” But later, David Holz of Midjourney is quoted as saying that Midjourney is trained on “a big scrape of the internet”.
Clearly there are bound to be copyrighted images included in any
”big scrape of the internet”. But do we know that “all the visual information in the system” is derived from copyrighted images? What about the rest of it? The detritus, the photos, the corporate images, the stuff from Pinterest, the IP whose statutes have expired, and the entire history of art and artwork that has a presence on the network? I’m thinking that the internet has become such a behemoth of cultural reference-making and generator of remixed forms that it might not make sense to go in with a copyright argument. Is it like: The internet is a vast wilderness, and the poplars have been included and don’t want to be.
Take proto-feminist 17th century painter Artemesia Gentileschi. I mean technically, the actual photograph of Artemisia Gentileschi’s Judith Slaying Holofernes would likely be copyrighted by the Uffizi in Florence, as the painting is part of their collection. (Link to the 20 gazillion reproductions of this painting that are available through a Google search and that would have been included in the Stable Diffusion training set). How do we feel about Judith being part of the big scrape? Does it not matter because she is dead, or does it not matter because the production of art in general is posited within historical discourse that contextualizes art in the context of other art? Is this whole copyright situation a replay of the early web when some people joined the HTML Writers Guild and hid their source code, and others released everything into the wild, with the thinking that their style and their work will be more widely seen, and widely known?
Which brings us to the question of influence, and how artists draw from and influence each other and always have. In addition to the influence of biblical and mythological stories relating to Judith, Gentileschi was also heavily influenced by Caravaggio, in this case not least because her painting is a direct revisitation of Caravaggios’s Judith Beheading Holofernes depicting her own version of the scene. The influences of Gentileschi, Caravaggio and the full roster of Baroque-era dudes flows all the way from the 16-somethings into work that is done today.
Polish artist Greg Rutkowski, well known in some communities for fantasy illustrations in popular games such as Dungeons & Dragons, has become a face of the debate about the potential for AI to displace the artist: MIT’s Technology Review reports that Rutkowski’s name is among the most used prompts on Midjourney and Stable Diffusion, having been input 93,000 times. So that is a huge, direct, and quantifiable mode of influence and I get it why he’s asking questions about copyright.
But in an argument made by nobody, can we trace for example stylistic references to Baroque painting - characterized by the use of “contrast, movement, exuberant detail, deep colour, grandeur, and surprise to achieve a sense of awe” - in the work of Greg Rutkowski and others who work in the fantasy genre?
Can’t resist featuring the endless creative fails of the machine, while we’re at it.
From the other side of left field comes another thing that happened this week: Artist Ben Moran was banned from Reddit because the work he uploaded was judged to look too much like an AI generated image. When challenged by Moran with evidence that he made it himself, a moderator from the subreddit replied, “Even if you did ‘paint’ it yourself, it’s so obviously an A.I.-prompted design that it doesn’t matter. If you really are a ‘serious’ artist, then you need to find a different style because A) no one is going to believe you when you say it’s not A.I., and B) the A.I. can do better in seconds what might take you hours.”
Still parsing the full meaning of this as a criticism!