Annie Murphy in a scene from Black Mirror episode Joan Is Awful, about an AI-generated reality show of the same name.Netflix
Hollywood currently has 11,500 reasons to be leery of the bullet train that is artificial intelligence (AI), and soon they may have 160,000 more. Those numbers refer to the 11,500 Writers Guild of America (WGA) members who have been on strike since May 2, and the 160,000 Screen Actors Guild (SAG) members who may be joining them on picket lines as of July 12. Meanwhile, as demonstrated by AI-centric series including Content Farm (CBC Gem) and the first episode of Black Mirror season 6 (Netflix) – “Joan Is Awful,” about an AI-generated reality show of the same name – the question of who (or what) controls image and content production has never been more urgent.
When the screenwriters first hit the streets, negotiations were focused on how the streaming giants (including AppleTV+, Prime and Netflix) have upended the traditional payment model, turning writers into gig workers and leeching away their security. But as the weeks march on, the guild is shifting attention to the threat that studios, production companies and the Alliance of Motion Picture and Television Producers (AMPTP) will use AI-generated material – which should be copyrighted, but is spit out by programs like ChatGPT-4 without credit or payment.
This isn’t happening on any scale yet, but as AI becomes more sophisticated, the lure will be greater. So the WGA wants a rule now, before it’s too late, that AI can’t be used to write or rewrite scripts, and that bots can’t be covered by bargaining agreements.
The AMPTP, however, won’t discuss it. They want to keep their options open. Filmmaking is staggeringly expensive because it requires scads of human artists who need to be paid, so anything that can cut those costs would be catnip to the money folks. Even as the writers picket, the studios are taking pitches from tech startups such as Runway, who promise to make entire films and series without all those pesky people. Their slogan: “No lights. No camera. All action.”
That’s why “Joan Is Awful” hit so hard. A tech executive (Canadian actress Annie Murphy, Schitt’s Creek) goes about her day, then comes home and flips on Streamberry (which uses the same opening buh-bom sound and coloured stripe logo as Netflix, wink wink), to find that her life is suddenly a series starring Salma Hayek, unfolding in real time. Only it’s not really Hayek – the actress licensed her face to an AI-generated performer. (Which is one of the IRL fears animating the actors’ contract negotiations.)
Both characters consult lawyers, and both are told there’s nothing to be done – everything was spelled out in the fine print they agreed to. The corporation is free to use what Hayek’s character calls “this deepfake heretic abomination” to shoot, edit and package infinite programs, even though the Streamberry CEO admits, “We barely know how it works.”
The episode dropped on June 16; a week later, online searches for “Netflix terms and conditions” had spiked by 1,524 per cent, according to independent reviewer casinoalpha.ie., and sites like Mashable were highlighting the “most Streamberry-ish” parts of Netflix’s actual agreement. Examples: Netflix can test user interfaces and promotional features on you; they can use your feedback however they want; they aren’t liable for any injuries or damages whatsoever; and disputes can only be settled in arbitration or small claims courts.
Cleverly, Netflix doubled down in their promotion: They invited ordinary users to be part of a “You Are Awful” Streamberry Virtual Experience. In a nine-page terms and conditions document, volunteers were told that Netflix could use any of their identifying information, including their likeness, home address and social media accounts, in any way they wanted, even on public billboards. And still, when billboards popped up in England reading “Shahab Is Awful” and “Nicholas Is Awful,” volunteers were shocked. (One wag sent a photo of his dog, Moriarty, and Netflix used it, too.) As the Streamberry CEO says in the episode, keeping viewers in “a state of mesmerized horror really drives engagement.”
Two people who saw this coming: Lauren Gillis and Alaine Hutton, who for 12 years have been co-artistic directors of Toronto’s Lester Trips theatre company (named after a robot who fell off a staircase during a tech demo). Their first-ever digital series, Content Farm, is a series of connected sketches that uses AI deepfakes – one totally fooled me – to ask the question, if you were a bot, what would you learn about humans, based solely on their internet behaviour?
It touches on content-scraping bots that steal identities to manufacture clicks, autonomous revenue maximizing systems programmed to self-destruct if no revenue is generated, and the pantomime of proscribed internet apologies. There’s also a dating series where two bachelors prefer a deepfake bachelorette to her real self. The duo wrote it three years ago, but it arrived on CBC Gem on March 29, just as ChatGPT-4 was launched.
“We all know that Instagram followers and social media commenters are bots, yet we don’t behave as if that’s the case,” Gillis said this week in a joint video interview with Hutton.
“The reality of what’s happening – that bot activity is growing exponentially, that it’s stealing our identities – is too difficult to process, so we ignore it,” Hutton agrees. Their three-episode series is deliberately cringy, she continued, built on the idea “that public mortification might happen to you tomorrow, yet you still enjoy the schadenfreude that’s it’s not you today.”
Interestingly, both Content Farm and “Joan Is Awful” (directed by the Canadian writer/director Ally Pankiw) star women, whose likenesses and bodies are particularly dehumanized and commodified by AI. “Women on screens are often just a series of gestures and skin textures,” Gillis said: eyes, lashes, hair, breasts, augmented and filtered beyond what looks “human.”
Yet bots aren’t inherently thieving, racist or misogynistic, she added – they’re just capitalizing on the theft, racism and sexism that humans already promulgate. Look at The Idol (HBO/Crave), where vast teams of alleged support staff make gobs of money by tooling and retooling the image of pop star Jocelyn (Lily-Rose Depp) – and where the series’ creators, including Sam Levinson (Euphoria) and the musician/actor The Weeknd, get to have their porny exploitation of Depp’s body and decry it, too.
“AI versus respecting people’s art and work, that’s a human problem,” Hutton said. “People who control content are constantly stealing already. With added tech it will be easier to edge creators out. The people at the top don’t care about the human casualties, and that attitude frightens me more than any tech.”
After all, you know what the fictional Joan Is Awful series doesn’t need? Writers.
Sign up for The Globe’s arts and lifestyle newsletters for more news, columns and advice in your inbox.