How to survive in pop? Create a deepfake version of yourself

FKA Twigs wants her AI avatar to interact with her fans while she makes music in the studio
FKA Twigs wants her AI avatar to interact with her fans while she makes music in the studio - Redferns

“In many ways for us it’s helpful. It was helpful when we did Now and Then. John is there because of AI,” said Ringo Starr last week, referring to former bandmate John Lennon’s Artificial Intelligence-enhanced voice on a track rescued from a grainy demo cassette and released as “the last Beatles song” last November. “So we have to thank [AI] for that.”

Starr was answering a question I’d put to him about technology and the future of pop stardom on a transatlantic Zoom Q&A session to promote his new EP, Crooked Boy. He’s well aware of the alarm that AI generates. “I know there’s a lot of fear out there that it’ll steal your soul,” Starr said, “but I won’t steal anybody’s soul, so you’re OK.”

The fact that the 83-year-old former drummer of The Beatles is extolling the virtues of AI is proof of how deeply the technology is penetrating the music industry. But using AI to digitally spruce up vocals – as in Lennon’s case – is nothing compared to the latest technological frontier that is opening up: using AI to create talking clones of musicians so they can interact with their adoring fans. Some experts believe so-called “deepfake” stars represent the future of pop fandom. And the industry is getting excited about it.

This brave new world is already upon us. Last month Cheltenham-born musician FKA Twigs said that she had created a deepfake computer-generated version of herself using AI. She said the deepfake, which she dubbed “AI Twigs”, is trained on her personality and can use her exact tone of voice to speak in many languages including French, Korean and Japanese. From later this year, the Grammy-nominated singer will use her AI Twigs avatar “to extend my reach and handle my online social media interactions, whilst I continue to focus on my art from the comfort and solace of my studio”.

FKA Twigs isn’t some wacky outlier, and her avatar isn’t just some publicity wheeze. Record labels are taking talking bots seriously. The owner of one independent UK record label believes that all labels are actively looking at how AI can increase fandom, including the “big three” labels of Universal, Sony and Warner Music.

“There are people sitting in major labels looking at this. FKA Twigs will be the first of many,” they speculate. “Some artists won’t do it because they don’t want to. But if you’re a fan of Billie Eilish or Harry Styles and you want to know what they had for breakfast, there’s a degree of detail that can be passed on using alternative means such as AI. The artist themselves may not have the time.”

K-pop band Aespa already have deepfake versions of themselves
K-pop band Aespa already have deepfake versions of themselves

Insiders argue that talking bots are an obvious progression in artist-fan relationships. “Pop fandom used to be run by a club which you’d join and you’d get sent stuff in the post,” says Guy Gadney, the chief executive of Charisma, an AI company that creates characters capable of having realistic conversations. Deepfake pop stars on social media are the “natural next step” of this, Gadney says; a 21st century equivalent of sending a stamped addressed envelope to an address in the back of Smash Hits and getting a Simon Le Bon pin badge back in return.

Some music industry insiders believe FKA Twigs-style deepfakes will not only change fandom – but will become integral to being a pop star full stop. “Musicians have been getting exponentially busier,” says Mark Mulligan, media analyst at Midia Research. “In the past, record labels did most of the marketing on behalf of artists. Now, artists have to have an authentic voice when they’re on social platforms, most of which are now video-centric.” The obvious solution? An AI bot. “A good AI bot, as long as people know it’s AI, can serve a perfectly decent role.”

This all sounds like J G Ballard-style dystopia. But it’s not. Here is how deepfakes work. A computer programme automates what someone looks and sounds like by scouring the internet and absorbing every video, photo and audio clip of that person. Once “trained”, the software then mimics and anticipates how the target person acts so it can listen and talk back. The movement is already big among Korean K-Pop bands. All four members of girl-group Aespa, for example, have virtual AI representations of themselves. The eight appear in music videos together. Aespa (the real version) are big business: they’ve played Coachella and modelled for Givenchy.

FKA Twigs testifying to the US Senate Judiciary Subcommittee on Intellectual Property
FKA Twigs testifying to the US Senate Judiciary Subcommittee on Intellectual Property - Shannon Finney

Of course, avatars in music became a talking point in 2022 when Abba launched their Voyage hologram show in east London. But in Voyage the so-called Abba-tars on stage are pre-programmed rather than spontaneous; you couldn’t have a conversation with them, unlike this new breed. Goldman Sachs recently launched its annual Music in the Air report, its flagship delve into music industry trends. The investment bank reckons that the existential threat posed by AI-generated music – fake songs – has subsided due to the low quality of tracks and the industry cracking down. Instead, Goldman believes that AI’s value lies in artists themselves using it for, among other things, “marketing purposes to grow [their] audiences”. In other words, AI Twigs-style bots.

And growing an artists’ audience is crucial to labels’ profits. As income from streaming plateaus, labels are looking at other ways to make money. One strategy lies in harnessing fandom. So-called “superfans” will be worth $4 billion a year to the industry by 2030, reckons Goldman Sachs. Nearly a quarter of music lovers classify themselves as superfans, according to music industry body the BPI. One way to these fans’ hearts, and wallets, is via merchandise, but another way of monetising these loyal fans is through technology.

The rise of online chatrooms is one strand of this. On apps such as Discord, artists can post updates directly to their followers. Sussex-born pop star Maisie Peters, for example, has 20,316 “members” on her Discord forum and she regularly posts messages to them. Artists can also send new songs to fans at the touch of a button; it’ll ping on their phones. Talking avatars are a logical progression in harnessing superfans.

There’s another pressing reason behind the rise of deepfakes: ownership. FKA Twigs, real name Tahliah Debrett Barnett, claims she is launching her bot because she fears that unscrupulous AI practitioners will be able to illegally replicate her likeness and “falsely claim my identity and intellectual property… threaten[ing] to rewrite and unravel the fabric of my very own existence.” The very “stealing your soul” that Ringo was talking about. By launching AI Twigs now, Barnett has seized the initiative. The fact that she made her comments in testimony to the US Senate Judiciary Subcommittee on Intellectual Property in Washington DC shows just how serious the situation is.

In the UK, the All-Party Parliamentary Group on Music recently called on the government to regulate artificial intelligence. The parliamentary group’s chair Kevin Brennan MP wants humans to govern AI before AI governs us. “Our central insight must always be that AI can be a great servant but would be a terrible master,” Brennan said.

But what about believability? Do people know, or care, that they’re talking to a computer-generated clone? “What we’ve seen is that people believe the characters they speak to much more than you might think they do. We suspend our disbelief pretty easily,” says Gadney, founder of Charisma, the AI tech company.

However Tom Stabb, a former digital marketing manager at Atlantic and Columbia records, believes the technology is not quite there yet. “It’s fascinating [but] with AI tools like that there’s always going to be that ‘uncanny valley’ of it not being quite 100 per cent right. But that doesn’t mean that it won’t be in five years’ time.”

There’s also a backlash brewing. Alex Connock is a senior Fellow at the Saïd Business School, Oxford University, and a specialist in AI and the entertainment industry. His research has unearthed some anti-avatar sentiment among fans. “The more synthetic content we create, the more people are craving the touchable live experience,” Connock tells me. “People want to feel the muddy fields of Glastonbury or the searing desert heat of Burning Man, precisely because those experiences are analogue, human and non-synthetic.”

And the one thing that AI can’t recreate is the churning mud of a festival field. Our souls are safe from a complete robot takeover – for now.