Troubled by the public stance that Stability AI began using copyrighted material to train its generative AI product, the company's head of audio, Ed Newton-Rex, has announced that he is leaving the company. The move sparked a broader discussion about the concept of “fair use” and the use of content without the creator's permission in AI models.
“I have resigned from leading the audio team at Stability AI because I disagree with the company's view that training generative AI models on copyrighted works is ‘fair use,'” Newton-Rex wrote on Twitter.
Reaction was mixed, with some praising Newton-Rex's position.
“Thank you for taking a stand and doing the right thing by giving up your composure,” wrote Reid Southern. “If we had more people like you working in machine learning and generative AI, we probably wouldn't be in this mess.”
However, others have questioned the definition of copyright infringement.
“So anyone who reads a book, listens to music, watches a movie, etc., is motivated by copyright infringement?” wrote Twitter user John Harvey.
The case, Newton-Rex explained, came when Stability AI responded to a US Copyright Office request for comment on generative AI. The AI developer has submitted a 22-page statement that includes the view that generative AI is a socially beneficial use of existing content protected by fair use and advances the goals of copyright law.
Stability is a focus of Newton-Rex Post, but the issue is bigger than any other company, he told Decrypt.
“My objection here is not really against Stability because Stability is taking the same approach that many other creative AI companies in the space are taking,” Newton Rex said. “What I'm really against is the cross-industry position. In fact, I have resigned from a group of companies that follow a similar approach.
Copyright refers to the legal right of creators to control the use of their work. Fair use, on the other hand, allows limited use of copyrighted material for education, evaluations, or research. In a perfect world, these concepts lead to a balance between the creator's ability to use and exploit their content and the public good.
In October, a group of authors including John Grisham and Game of Thrones creator George RR Martin joined a lawsuit against ChatGPT creator OpenAI, alleging that their work was based on training data for popular AI models. That same month, a similar lawsuit against Midjourney, Deviant Art, and Stability AI was struck down by a federal judge who said the plaintiffs did not provide enough evidence to support their copyright infringement claims.
Newton Rex says the Stability AI audio model was trained on music licensed from the digital music library AudioSparks.
“[AudioSparx] It has a great collection of music, and we've done revenue share, with the philosophy that if our model works well, they'll do well. “That's the idea, and that's one way I think it could work.”
“I don't think generative AI and creative industries should be enemies,” Newton-Rex continued. “You can work together.”
But while he's optimistic that creators and AI developers can get along, he acknowledges that artists have every right to be concerned about how text, music and other media are used to train AI models.
Newton Rex reiterated that Stability AI's Stable Diffusion was trained on publicly available data. However, he noted that the use of databases such as the LAION-5B is common in the industry; This may include material ripped from the Internet that has been copyrighted, he said.
“In the industry as a whole, image generation seems to take a pretty permissive view of other methods and data, certainly text generation, of what they're trained in, which I don't agree with,” he said.
While Newton Rex said he didn't have specific plans when he left Stable AI, he said the discussion around copyright and fair use is important.
“I think this is a very important conversation and people are talking about this copyright issue because I think we have to do it and we have to decide on it.”
Stay on top of crypto news, get daily updates in your inbox.
