Like most human endeavours, art is no stranger to automation. But in the world of digital art, this automation is doing more harm than good.
Ever since humans have had computers they have tried to get them to create art, as seen in projects such as Earl Birney’s Space Conquest: A Computer Poem or the writings of Mindscape’s Racter. Recently, there have been several programs released to the public that allow users to commission an AI to create pictures for them, though they’ve come into some controversy.
The AI art program wave started with Craiyon (formerly known as DALL-E mini) this past summer. Users are invited to input a prompt and the program will create nine pictures based on how it interprets that prompt. People would then generally post these pictures online, highlighting the most bizarre ones.
This was less of an artistic revolution and more of a toy for people online; most prompts would be along the lines of “George W. Bush as Thanos” or “Guy Fieri giving a TED Talk lecture.” Headlines at the time such as Wired’s and Science Focus’s reflected this, focusing on Craiyon’s capacity for meme-making rather than any potential for making traditional art. While some speculated that the potential was there, Craiyon never went above the memes.
Since then, other programs have sought to improve upon Craiyon’s sticking points. OpenAI, whose code Craiyon was built upon, later released DALL-E 2, promising more realistic image generations. They have been joined by numerous others, such as WOMBO Dream, Night Cafe, and StarryAI, all with similar promises.
The posts generated by these sites and shared by users are much different from Craiyon’s pictures. They move away from the meme-style prompts and create more original pieces. However, their originality has been put into question.
Greg Rutkowski is a digital artist who specializes in making fantasy landscapes. His work has been featured in multiple large franchises, such as Dungeons and Dragons and Magic: The Gathering, but it also has attracted many of those using AI art programs. One feature of these apps is the ability to input an artist’s name in the prompt, and the program will mimic that artist’s style. Rutkowski’s name was a popular prompt, and soon there was art appearing under his name that he never drew.
“It’s been just a month. What about in a year?” he said in an interview with MIT Technology Review. “I probably won’t be able to find my work out there because [the internet] will be flooded with AI art… For me and many other artists, it’s starting to look like a threat to our careers.”
Beyond just aping the style of human artists, these programs have been used to outright steal artist’s work as well. Korean graphic designer HAJE714 was live streaming while working on a piece, and a viewer took a screenshot of their stream. This viewer uploaded the screenshot to the program Noble AI, altered the picture slightly and then uploaded it to Twitter as though it was their own work.
In a follow-up Tweet, HAJE714 clarified why this was an issue: while in this instance the plagiarist was caught, there is not much deterring other users from doing the same, with a lack of legal repercussions. This time, the plagiarist was caught in the act, but with how large the internet is, there is a sizable chance that artists won’t be able to catch everyone who does this.
While not every user of these programs does so to undermine artists, the issue is that there is nothing in these programs to prevent users from doing so. Users are encouraged to create art in the style of various artists, regardless of whether they get approval or any kind of formal consent.
Additionally, the ability to alter existing pictures in these programs quicker than one could do so in programs such as Photoshop makes it possible to steal an artist’s work-in-progress and claim it as your own. And since the AI can finish a piece faster than a human, the stolen piece can be published before the original, making plagiarism claims harder to verify.
While much of the online discussion about these programs revolves around the increases in their pictures’ quality, these improvements only serve to heighten the moral issues within them. These programs blur the line between original content and theft, as while the pieces generated purely by the programs are original they borrow heavily from existing artwork and art styles. As well, the programs make art theft much easier, even if that is not their primary purpose.
So far, none of the major programs have made any comments on these issues, let alone indicate how they will attempt to resolve them. But given how reliant these programs currently are on existing artists, these issues will likely need to be resolved before these programs can achieve any mainstream success.