Why This Online Archivist Isn’t Feeling Much Angst About AI-Generated Art

The rise of the creative machines–AI routines that can generate original pictures in response to simple descriptions of the desired image–isn’t something to fear, according to a longtime scholar of digital culture.

“I am no more scared of this than I am of the fill tool,” Jason Scott said in a talk at The Atlantic Festival(Opens in a new window) in Washington, comparing AI image generators like DALL-E 2 and Stable Diffusion to features in Adobe Photoshop. “Or the clone brush.” 

Scott, an archivist and curator at the Internet Archive(Opens in a new window), treated the audience to a slideshow of AI illustrations drawn for such requests as “a lion using a laptop, in the style of an old tapestry” or “Godzilla at the signing of the Declaration of Independence.” The results generated by these tools, all trained on vast databases of published art, looked both silly and showed a marvelous level of detail.

Later, he staged a live demo of DALL-E. It ably fielded a request for “the moment the dinosaurs went extinct, in the style of Art Nouveau,” but the images it generated for “people at the Atlantic Festival having a great time, in the style of Edward Hopper” evoked yacht rock instead of a Washington magazine’s conference.  

“I can’t get angry at this toy,” Scott told his onstage interviewer, Atlantic executive editor Adrienne LaFrance–warning a moment later that “this toy won’t stay neutral.”

His concern isn’t that these tools will snuff out human creativity or put human artists out of work–although he did cite examples of people using AI image generators for such repetitive real-world tasks as creating textures for video games. 

Instead, Scott fears that today’s AI angst will lead to a replay of past instances of creators of remix-based art being at least scorned by the industry’s gatekeeping types, at worst finding themselves on the wrong end of copyright litigation. 

“It’s also like sampling,” he said, noting that he’s already seeing services that can tell artists if any of their published work has been ingested by an AI image generator’s machine-learning model. “There are some communities saying we don’t want any of that here,” Scott said. For instance, last week Getty Images banned the upload and sale of AI-created pictures(Opens in a new window).

In a conversation after his talk, Scott expanded on those themes.

“This purely is a tool,” he said after scrolling through a vast array of Discord channels for the AI-creation tool Midjourney(Opens in a new window) on his phone. But, he added, it’s not like other artistic tools in that it needs fuel, in the form of those vast training databases of images: “A pencil needs graphite, but it doesn’t need previously drawn pictures to draw.”

Scott suggested that clients determined to keep out AI art might impose a proof of work requirement: “You might eventually have a case where an artist has a camera over their drawing board and they have to show the five intermediate drawings when they submit a drawing to a company before the company will pay them real money.”

Then he offered a sales pitch from the future for a hypothetical artist out to make non-synthetic image creation part of their personal brand: “Don’t panic, organic’s here.”