Pin It
Black Mirror

Should we use AI to bring musicians back from the dead?

AI has helped create ‘new’ songs from the likes of Amy Winehouse and is now being used to produce a final Beatles song – but is it ethical to tamper with the legacies of late musicians?

Some artists are deemed immortal through the lasting influence of their music, but new technology seeks to actually raise their voices from the dead. Through emerging AI models, fans are reworking the vocals of deceased musicians into new songs: Freddie Mercury croons over George Michael's ‘Careless Whisper’, and Michael Jackson attempts ‘Wake Me Up’ by Avicii. But these AI models aren’t merely imagining a universe where the musical greats lived on and dedicated their autumn years solely to questionable karaoke: just last week, Paul McCartney announced that he had used AI to create the final Beatles song.

Fans were divided, with many envisioning an ethically ambiguous, robotic version of John Lennon’s voice singing lyrics written in 2023. In reality, though, AI had only been used to extract Lennon’s vocals from an ageing tape gifted to McCartney shortly before his death. “We generally distinguish between analytic and generative uses,” explains Dr Tom Collins, lecturer in music technology at the University of York. Analytic AI, the more traditional form, is used to analyse existing data to aid in prediction or automation, meaning it can be useful for more manual tasks. Generative AI, on the other hand, is able to learn from data and then create new data from its findings. “In this case, the AI has just done something analytical, and humans then take that and do something generative with it.”

These ‘analytic’ forms of AI in music, though sounding eerily futuristic, have been around for decades. It’s generative AI, such as timbre transfer, that’s behind all the viral TikTok covers – and it’s seeing rapid development thanks to the emergence of faster, more powerful computers and greater access to musical data. So how does it work? “If someone’s got a lot of Queen on their computer, they can run that through the source separation algorithms to get their own homemade, isolated version of Freddie Mercury’s voice,” Collins says. “Then with that model, you can take in another single isolated vocal line from say, George Michael’s ‘Careless Whisper’, and show it to the Freddie Mercury model. What you get out is something that sounds like Freddie singing ‘Careless Whisper’.”

Keeping up with the accelerated evolution of AI is often headache-inducing, in part due to all the robot-gains-sentience-and-destroys-all movie tropes we’ve been raised on. With this in mind, it’s no wonder that AI is a divisive issue among artists. While Grimes gave her blessing to fans to use her vocals to experiment with the technology (with the caveat that they don’t generate a “Nazi anthem”), Tyler, the Creator has been explicit in his desire to not be reincarnated by AI after his death. Similarly, Ice Cube referred to ‘Heart On My Sleeve’, an AI-generated duet between Drake and The Weeknd, as “demonic” and urged Drake (who, by the way, in 2018 released a ‘duet’ that featured previously unheard vocals from Michael Jackson) to sue the creators. The song was later taken down by Universal Music Group.

While modern artists are granted the opportunity to provide their consent, musicians who died before generative AI became accessible didn’t get the chance, posing an increasingly urgent ethical dilemma. Creative AI researcher Kyle Worrall advises that the best way forward would be to contact families of the deceased artists to seek permission before using their vocals. “While this wouldn’t be as ethical as getting permission from the artist when they were alive, this seems like a fairer middle ground considering that it has been impossible to fully predict what AI will be able to do as it advances,” he says.

Fans experimenting with AI to create fake covers, however, adopt the mindset that it’s all done with pure intentions. Creators like 22-year-old YeezyBeaver, who went viral with his AI-generated audio of Kanye West singing ‘Hey There Delilah’, receives no financial gain from his creations and explicitly states it’s purely for entertainment and exploratory purposes. “It’s just really fascinating to me,” he says. “The speed at which technology is improving and becoming more accessible for the average person is crazy.” Still, he sees how the rise of AI can pose problems for musicians. “I can see the trouble though when it comes to someone using this tech and trying to pass it off as if they are the actual person,” he says.

While these fan-created covers seem relatively harmless, there’s growing anxiety that AI may allow the wider music industry to profit off artists even after they pass away. After all, exploitation of even living artists is rife: take TLC’s bankruptcy in the 90s, despite releasing a best-selling, Grammy-winning record; Kesha being trapped in a recording contract with her alleged abuser; or Britney Spears forced to tour the world and generate hundreds of millions of dollars in profit, while under a conservatorship that entirely removed her autonomy. Many labels already prioritise profit over the wellbeing of the artists they’re supposed to look after, so it’s hardly a leap to suggest that greedy execs could continue exploiting musicians even when they’re dead.

However, Worrall stresses that there are ongoing discussions of establishing laws and legislation to protect artists. “At the moment it’s a bit like the Wild West without consistent rules and regulations to outline what is allowed, but this is likely to change over time,” he says.

“[The AI] doesn’t have the same lived experiences as a human. It hasn’t fallen in love or watched a sunset. It hasn’t gotten drunk and had a hangover. So it’s probably going to be a little bit boring” – Nick Bryan-Kinns

There also seems to be a lack of public demand for entirely human-free music, as even YeezyBeaver admits that nothing will ever beat “the real deal”. After all, part of the lore surrounding legendary artists lies in who they were as a living, breathing person, and how that translates into their songs. “[The AI] doesn’t have the same lived experiences as a human,” Nick Bryan-Kinns, Professor of Interaction Design at Queen Mary University of London explains. “It hasn’t fallen in love or watched a sunset. It hasn’t gotten drunk and had a hangover. So it’s probably going to be a little bit boring.” That‘s not to say AI has no place in posthumous releases – it’s just more likely to exist as a creative assistance to production, rather than as a ‘replacement’.

Still, Gen Z’s use of the technology hasn’t gone unnoticed, with Bryan-Kinns admitting their penchant for reviving dead artists through unhinged mashups was a pleasant surprise to scientific communities. “Most scientists who come up with these AI models wouldn’t imagine it going that way, and that’s the great thing,” he says, though admittedly aware these covers are likely a novelty that will wear off. “I think what will be interesting then, is how does it change music-making practice and people’s processes? And does it lead to a new genre of music? That’s the potential.”

Maybe if we begin to view AI as a tool, rather than a replacement for human creativity, it will become an era-defining musical breakthrough. After all, 50s soul musicians couldn’t possibly have fathomed a world where their songs would be given new life as samples over hip-hop beats, just as Freddie Mercury likely didn’t predict his vocals would be plucked from his recordings and deposited into pop songs 30 years after his death. The way we deconstruct and reinterpret the art of our predecessors is part of the endless dialogue music allows us to create: perhaps AI is simply a new chapter, and the pervasive negative reactions to it are just part of the process too – as after all, almost all technological advancements in music, from autotune to the electric guitar, faced backlash at one point. 

As Grimes poignantly put it when defending the use of AI in music: “art is a conversation with everyone that’s come before us”. It just so happens that the current topic of conversation is ‘What would it sound like if Kurt Cobain was on the Shrek soundtrack?’.

Join Dazed Club and be part of our world! You get exclusive access to events, parties, festivals and our editors, as well as a free subscription to Dazed for a year. Join for £5/month today.