Should artists be paid for training data? OpenAI’s VP wouldn’t say

Should artists whose work was used to train genetic artificial intelligence like ChatGPT be compensated for their contributions? Peter Deng, vice president of consumer products at OpenAI — the maker of ChatGPT — didn’t want to answer when asked on the SXSW main stage this afternoon.

“That’s a great question,” he said when SignalFire partner (and former TechCrunch writer) Josh Constine, who interviewed Deng at a large gathering, asked the question. Some of the crowd of spectators shouted “yes” in response, which Deng acknowledged. “I hear from the public that they do. I hear from the public that they do.”

That Deng avoided the question is not surprising. OpenAI is in a delicate legal position when it comes to the ways it uses data to train productive AI systems, such as its DALL-E 3 art creation tool, which is built into ChatGPT.

Systems like DALL-E 3 are trained on a huge number of examples—artwork, illustrations, photographs, and so on—usually sourced from public websites and datasets on the Web. OpenAI and other AI makers argue that fair use, the legal doctrine that allows the use of copyrighted works for secondary creation as long as it is transformative, shields their practice of taking public data and using it for education without compensation or even credit artists.

OpenAI, in fact, recently argued that it would be impossible to build useful AI models without copyrighted material. “Training artificial intelligence models using publicly available materials on the Internet is fair use, as supported by long-standing and widely accepted precedent,” the company wrote in a January filing. suspension. “We see this principle as fair to creators, essential to innovators, and critical to U.S. competitiveness.”

The creators, unsurprisingly, disagree.

A class-action lawsuit brought by artists including Grzegorz Rutkowski, known for his work on Dungeons & Dragons and Magic: The Gathering, against OpenAI and several of its rivals (Midjourney and DeviantArt) is headed to court. Defendants argue that tools such as DALL-E 3 and Midjourney reproduce artists’ styles without the artists’ express permission, allowing users to create new works that resemble the artists’ originals for which the artists receive no payment.

OpenAI has licensing agreements with some content providers, such as Shutterstock, and allows webmasters to block its web crawler from scraping their site for training data. Additionally, like some of its rivals, OpenAI allows artists to “opt out” and remove their work from the datasets the company uses to train its image-generating models. (Some artists have is described the exclusion tool, which requires removing a single copy of each image along with a description, as cumbersome, however.)

Deng said he believes in artists must they have more power in creating and using artificial intelligence production tools like DALL-E, but it’s not certain, exactly, what form that might take.

“[A]artists should be a part of it [the] ecosystem as much as possible,” Deng said. “I think if we can find a way to make the flywheel of art creation faster, we’ll really help the industry a little bit more… In a sense, every artist has been inspired by artists who have come before them, and I wonder how much of that will be quickened by it.’