The artists, in an open letter directed at OpenAI said; ‘we are not your: free bug testers, PR puppets, training data, validation tokens’.
OpenAI’s unreleased Sora, an artificial intelligence (AI) text-to-video generating model was leaked by a group of artists yesterday (26 November). Although, access to the leaked AI model has since been revoked.
The ChatGPT-maker first shared details on Sora in February. While it wasn’t made available to the public, some visual artists, designers and filmmakers got early access to the model to give OpenAI feedback.
However, yesterday, the artists, in protest of being OpenAI’s “unpaid R&D,” leaked the model and accused the company of using artists as a means of PR.
In an open letter signed by nearly 600 signatures on Hugging Face, a public repository of AI models, the group said they received access to Sora with the “promise to be early testers, red teamers and creative partners”.
However, they claimed that they were instead “lured into ‘art washing’ to tell the world that Sora is a useful tool for artists”.
The group said that hundreds of artists provided OpenAI, a company valued at more than $150bn “unpaid labour through bug testing, feedback and experimental work,” with a few artists receiving the opportunity through a competition to screen their Sora-created films.
The group claimed that this was “minimal compensation,” ‘paling’ in comparison to the “substantial PR and marketing value OpenAI receives” and said that every output they created needed to be approved by OpenAI’s team before they could share it.
While the group behind the leak clarified that they are not against using AI for art, they claimed to not agree with how the Sora program is being rolled out.
In March, OpenAI released some videos created by visual artists and directors using Sora, which included a short film titled Air Head, made by Toronto-based multimedia production company and indie pop band Shy Kids.
“Since we introduced Sora to the world last month, we’ve been working with visual artists, designers, creative directors and filmmakers to learn how Sora might aid in their creative process,” the company wrote in a blog.
TechCrunch, who initially reported the leak said that any user can generate 10-second videos up to 1080p resolution by typing a short text description.
Meanwhile, The Verge reported that OpenAI would not confirm whether the “alleged Sora leak” was authentic or not.
Last year, OpenAI released Dall-E 3, its upgraded text-to-image generating model able to understand “significantly more nuance and detail” than its previous iterations.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
