Deepfake Ryan Reynolds trying to sell you a Tesla is a lot like what the future of Hollywood might look like. “How much do you think it would cost to own a car that’s this fucking awesome?”
The actor asks in a video that has since been taken down but was widely shared on Twitter. He is wearing thick black glasses and his mouth moves independently of his face.
On a measure of how real it looked, the video, which started going around last month, scored as very fake. Then, Kevin Paffrath, a YouTuber who gives financial tips, said that he had made it as a way to get Elon Musk’s attention. (Which is exactly what happened: the CEO of Tesla responded to Paffrath’s tweet with “nice.”) People asked Reynolds to sue in other places on Twitter. Instead, his production company made a similar bad video in which a gray-looking Elon Musk promoted gin made by Aviation, a company Reynolds co-owns. This movie has also been taken down since then.
“Finance guy sucks up to Elon Musk on Twitter” isn’t exactly earth-shattering news, but the exchange shows a much bigger problem: AI is making it possible for anyone to deepfake famous faces into any video they want. And artists, for their part, are becoming more and more aware of how AI could affect their work. Since the Writers Guild of America is already on strike, in part because of a similar danger, the actors’ union and studios are likely to bring up Fake Bruce Willis and Fake Ryan Reynolds as the latest steps toward a future where AI rules.
Duncan Crabtree-Ireland, executive director and chief negotiator for the Screen Actors Guild—American Federation of Television and Radio Artists (SAG-AFTRA), says that the hype around the technology means that it will be a focus of the talks. This is especially true since contracts are only negotiated once every three years. “It’s hard to even imagine where AI will be in three years, given how far it’s come in the last 18 months,” he says.
In a message asking its members to approve a strike, the group said it wanted a contract that would protect its members from losing money because of “unregulated use of generative AI.” Monday, June 5, is the date, and on June 7, SAG-AFTRA will start talking with the Alliance of Motion Picture and Television Producers (AMPTP), which is the group that speaks for the studios. It would be the first time since 2000 that performers went on strike.
Since the days of grainy sports video games, SAG has been worried about machine learning tools. Back then, the guild was worried about how easy it was for game companies to put pro athletes into Madden games. Now, companies in Hollywood are making Harrison Ford look younger and recreating the voices of people who have died.
In light of this, it’s not difficult to picture a scenario in the future in which a wide-eyed actor signs on for one season of a vampire TV show and then, two seasons later, their AI substitute bursts out of a coffin. They don’t get paid more in the meanwhile, even if the AI-generated character was based on their appearance and performance.
According to Crabtree-Ireland, “the nature of the impact on performers is unique, especially with generative AI tools that can be used to recreate a performer’s image, likeness, or voice persona or to do things that they hadn’t initially considered ever doing.” “That’s a problem.”
The right of publicity, commonly referred to as name, image, and likeness rights, shields actors, like all Americans, from being used for commercial advantage by others by using their likeness. By include “informed consent” in future contracts, SAG hopes to strengthen these rights and stamp out predatory clauses like the vampire example: According to the union, some uses of AI require disclosure and compensation.
The same cannot be said for writers’ reliance on publicity rights. When large language models, or LLMs, scrape someone’s work, if they have the rights, they may be able to seek redress or compensation—but only if the new work is regarded as a copy or derivative of the original script. According to Daniel Gervais, a professor of intellectual property and AI law at Vanderbilt University, “this is not very likely if the AI has learned from hundreds of scripts or more.”
Furthermore, talent agents are concerned with this scraping of performers. Leigh Brecheen, an attorney who specializes in the entertainment industry, claims that her biggest concern is that her clients’ prized traits will be taken in an imprecise manner. Imagine a producer summoning a digital performance with Denzel Washington’s intense piercingness but completely avoiding paying him. According to Brecheen, “the majority of negotiated contracts for on-camera performers will contain restrictions against the use of name, likeness, or performance in any work other than the one for which they are being hired.” I also object to the studio using the performance to develop AI. It is necessary to redefine AI work as an amalgam of numerous humans, as Crabtree-Ireland notes, in order to address this.
But will it matter to people that what they’re watching was created by an AI trained oh Human Scripts and performances? Unprotected writers rooms for police procedurals or sitcoms would likely become less necessary when ChatGPT and other LLMs can generate filmable sequences based on straightforward suggestions. Real danger also exists for voice actors, especially for those who haven’t already achieved fame for their on-camera appearances. Hany Farid, a lecturer at the University of California, Berkeley who specializes in evaluating deepfakes, claims that voice cloning is largely a problem of the past.
In the short run, the majority of AI-generated performers might come across as very unlikeable, like Fake Ryan Reynolds. It seems more plausible that people will accept artificial intelligence (AI)-produced audiobooks or a digitally recreated Darth Vader voice rather than a movie supported by the chiseled shoulders of an AI-sculpted action hero in the style of GigaChad.
However, in the long run, spectators may not care whether the performer in front of them is human if AI replicants succeed in overcoming the uncanny valley. Professor of law and artificial intelligence at Emory University Matthew Sag adds, “It’s complicated. “Writing work can be intruded into in a minor or progressive manner. The replacement of performers will probably be all or nothing.
The main topic of discussion when the actors’ union and Hollywood studios meet next week will be economic justice: According to the organization, it is become harder for guild members to “maintain a middle-class lifestyle.” Unions contend that there is a discrepancy in current times between the success of a movie or TV show and residual pay, as well as lengthier intervals between seasons that are becoming ever-shorter and result in less time spent working.
In this situation, AI might be Hollywood’s newest strategy for creating more content with fewer human creators. The entire situation would be unimportant if it weren’t for the fact that Reynolds was created by AI. Union strikes are still a possibility as a result. “They’ve got a 2023 business model for streaming with a 1970 business model for paying performers, writers, and other creatives in the industry,” claims Crabtree-Ireland. That is unacceptable.
