There’s a widely shared moment in the Studio Ghibli documentary The Kingdom of Dreams and Madness where animation visionary and famously bad father Hayao Miyazaki watches a demonstration of AI animation – well, more specifically, he’s just made its creators cry by calling their rendering of a crawling, zombielike corpse “an awful insult to life.”
One of Miyazaki’s peers asks the devastated developers to explain their ultimate goal, to which they respond, “A machine that draws pictures like people do.” There’s an awful pause. We cut to Miyazaki drawing alone. “I feel the world’s end is near,” he remarks.
The documentary turns a decade old this year, and Miyazaki’s dread couldn’t feel more relevant. In the wake of procedurally generated images like OpenAI’s DALL-E, NFTs being made out of stolen artwork, and the chance for movie fans to reimagine films in another director’s aesthetic, it seems like the intersection between artificial intelligence and art will be the next big thing or at least the next big sticking point for artists to rally against.
So far, AI has offered people without the resources of Hollywood studios the ability to demonstrate their own technical skill set, even arguing that they can produce better results for the industry’s visual effects houses. But the extent of AI-altering has stretched past films and into the lives of the actors starring in them for years.
AI speech generators offer the ability to mould any existing celebrity voice to say whatever someone wants them to say. While it’s not hard to realise when something’s been faked, that didn’t stop a slew of people abusing the technology for offensive results.
One ITV sketch show has even taken advantage of misrepresenting celebrity likenesses – Deep Fake Neighbour Wars intentionally brushes up against some defamation charges by imagining a slew of A-listers getting up to unlikely scrapes by living together. It looks like the real owners of the digitally-generated faces have decided legal action is not worth the fuss, probably because the show has been called “without question the worst television programme ever made.”
AI’s ability to create speech and likenesses seems like a case of the law being unable to catch up with new violations. What precise law has been broken by, for example, Zack Snyder fans deepfaking James Gunn to announce that, per their wishes, the Snyderverse has been sold to Netflix? On a less trivial front, the legal grey zone of AI tech has been exploited to incredibly serious and violating ends – with the UK only recently altering the Online Safety Bill to account for the creation and sharing of deepfaked pornography without people’s consent.
A human touch
What’s stopping networks and studios from utilising AI for any of the above purposes in their film and TV production? Nothing – they’ve already started. In 2021, Morgan Neville’s documentary Roadrunner, which celebrated the life of celebrity chef Anthony Bourdain, came under fire for deepfaking the late celebrity chef’s voice posthumously.
Neville addressed the controversy with a flippant, “We can have a documentary-ethics panel about it later,” and falsely stated that he had received permission from Bourdain’s ex-wife to reconstruct his testimony artificially. Nearly two years on, it’s perhaps time for that panel.
With its endless zeal for replicating the past, Disney has used AI to replicate the past for the past few years. Fans were ecstatic at the returns of original trilogy-era Luke Skywalker and Darth Vader across their slate of Disney+ series. But with Mark Hamill’s voice having aged since 1983 and James Earl Jones wanting to retire from his iconic role, they opted to use the voice cloning technology of Respeecher, a Ukrainian tech company, to replicate both actors’ recognisable voices.
What Disney did was by no means unlawful – Hamill was on set with Lucasfilm on The Mandalorian and The Book of Boba Fett, and Jones signed away the rights to his bassy tones for Obi-Wan Kenobi – but it sets a weird precedent. We see here the first steps in sidelining actors from their own performances as if characters could ever be distinct from or constructed without actors bringing them to life. Did Jones receive royalties for his “performance” in Kenobi? Can an AI voice join an actor’s union?
Of course, Hamill and Jones are incredibly successful and marketable performers; AI performances’ greatest impact will be on jobbing actors’ careers. Voice artists in Latin America’s expansive dubbing industry are already missing out on potential work because AI firms – who offer services for TV, film, non-fiction, and YouTube content – offer the cost-cutting alternative of not needing to pay a human being. It gets worse: desperate for pay, voice actors have been taking “poorly paid recording gigs at AI voice-over companies, training the very technology that aims to supplant them”.
Is AI art original?
This highlights the biggest contradiction of AI-created performances and art: they’re created off the back of existing audio-visual references, so claims of their originality are ludicrous. Take “AI Seinfeld”, officially titled Nothing, Forever, an endless stream of artificially generated Seinfeld content created by Mismatch Media, which became a hot ticket on Twitch earlier this year, giving viewers an uncanny-valley riff on a pop cultural icon.
Nothing, Forever is undoubtedly the product of a lot of labour, with its programmers acting almost like network producers of a weird, intentionally janky sitcom. How different is “AI Seinfeld” from the cynically imagined streaming reboots of dormant properties like Fuller House or Night Court? (Well, at least they’re not transphobic, which was how Nothing, Forever got banned from Twitch.)
Mismatch envisions a future where streaming can entertain without production costs or pauses. However, they haven’t yet explained how they will produce anything original – their big hit would not have been possible without training an AI with nine existing seasons of Seinfeld. On Respeecher’s website, they try to address what they think is “the most consistent criticism” of their AI work – a cloned voice doesn’t yet sound authentically human. The more pressing problem is being avoided: we haven’t yet imagined a use for AI entertainment that doesn’t seek to evade paying workers or exploit their labour.
As for its creative legitimacy, it’s sincerely unlikely that anyone invested in AI art other than for technological or financial reasons will be swayed by the artistic opportunities offered. Anyone truly invested in art is invested in the artist making it. Those who tout AI being the future have a hidden agenda: they need to convince artists to come aboard to improve their technology so they can eventually make money without artists, without people, and without labour.
This may surprise some, but the human effort and toil needed to create art isn’t a flaw but a feature – art is something defined as human beings interacting with and commenting on the world. I do not want to see AI creating art because I don’t think it’s art; every piece of AI artwork you have seen is just a recycling of someone else’s work. It is not puritanical or regressive to say that making art is an inherently human activity.
It may take a while for this to become the dominant opinion, though: Robert Zemeckis has started planning a film using the same deepfaked technology as a Tom Cruise impressionist on TikTok, and Twitter is full of pipedreams imagining new ways to centralise the viewer and not the artist’s work in entertainment’s future. Hey Netflix, I hear you’re looking for ideas…