DeviantArt Dreamup Launched: Artists are confused to learn that AI training will be done automatically with their work. DeviantArt, a website dedicated to sharing artwork, announced on Friday the launch of DreamUp, an artificial intelligence-driven text-to–image generating service built on the Stable Diffusion platform.
Concurrently, DeviantArt introduced a programme that theoretically allows artists to opt out of AI picture training, but also made everyone’s artwork opt-in by default, frustrating a lot of people. DreamUp is an innovative AI that generates the artwork in response to user-supplied language.
Ironically, some DeviantArt users might have issues with DreamUp because the company learned how to generate images by analysing hundreds of millions of images scraped from sites like DeviantArt and collected into LAION datasets without artists’ permission, thanks to its foundation in Stable Diffusion.
this is how dreamup is being developed. it adds training on non-opt-out artwork on top of the stability diffusion model, which is ALREADY trained using unethically sourced data. they CANT UNLEARN those parts.
this whole discussion is not in good faith unless they address this pic.twitter.com/KnQZerCIol
— svlt ✈️ (@svltart) November 11, 2022
Due to its web-scraping nature, Stable Diffusion sparked heated controversy earlier this year among artists who question the morality of AI-generated art, as we’ve chronicled extensively on Ars. There are artistic circles that are so opposed to AI-generated artwork that they outright forbid it.
It seems that DeviantArt is trying to appease artists who might be offended that their work is being used to train AI image generators in anticipation of a reaction. Artists who do not wish to have their images included in third-party image datasets can disable participation by checking a special “note” flag in their image settings. It is unclear, however, whether or not this flag will be respected by external picture scrapers.
Heads up, DeviantArt turned heel and as far as i can tell all of your existing deviations will have to be manually opted out of AI data sets.
YOU ARE OPTED IN AT DEFAULT
Extremely scummy move https://t.co/FcX03zvnI3 pic.twitter.com/S5eCAzq9gZ
— ɪʀɪꜱʜ ᴍᴇɴᴀᴄᴇ I Λ П | F Λ Y (@IanFayArt) November 11, 2022
Artists on DeviantArt will soon have the option to prevent their work from being used to train DreamUp, but only after submitting a form for human evaluation. Some members of DeviantArt have even threatened to remove their whole portfolios and cancel their accounts in protest of this policy.
DreamUp’s support page on DeviantArt’s website similarly takes a defensive stance, claiming that the site’s owners did not permit the use of their data by artificial intelligence image models like Stable Diffusion. In addition, the site aims to clear up some of the most widespread myths around artificial intelligence image synthesis farther down the page.
What about DreamUp by itself? We tried out the service and found it to be very similar to a standard Stable Diffusion setup. Each member of DeviantArt can try out the service with five free prompts, and extra prompt credits can be purchased with a monthly subscription to one of the CORE plans, which cost between US $3.95 and $14.95.
If you’re technically savvy enough to install a package from GitHub or have the iPhone app Draw Things, you can utilise Stable Diffusion locally for free.
Update: In response to community feedback, DeviantArt issued a change to their AI dataset policy a few hours after press time, declaring that all photos hosted on DeviantArt would be “automatically tagged as NOT allowed for use in AI datasets.”
Automatic opt-out of third-party AI datasets using the “noai” tag appears to be the extent of this at the present; this is a label that dataset scrapers will have to willingly honour in order for it to be effective (we know of no plans to honour this flag at this time). Artists who don’t want their work used in DeviantArt’s DreamUp AI training will still need to fill out a form.