How AI ‘Voice Fakes’ Are Changing Music

Since then, Paul McCartney announced that he would be putting out the “last” song by the Beatles. He used AI to extricate John Lennon’s vocals from a demo recording and completed the song more than 40 years after the Beatle’s death.

If you wanted to, you could listen to Kanye West singing “Hey There Delilah” by the Plain White T’s, or Harry Styles covering “Cardigan” by Taylor Swift, even though neither musician has recorded themselves singing those songs. You can even listen to former president Donald Trump singing “National Anthem” by Lana del Rey. It’s all thanks to AI.

Darden School professor Yael Grushka-Cockayne said that while the ubiquity of AI-generated art is new, it isn’t a novel application of the technology. (Photos by Dan Addison, University Communications)

“That you can create new works of art in the style of very well-known artists and that you can train models to come up with new kinds of creative pieces of work has been out there for several years,” Yael Grushka-Cockayne, a professor at the University of Virginia’s Darden School of Business, said. “But it’s now more accessible to everybody.”

That increasing accessibility has led to a flurry of activity. The Recording Academy even updated its rules to allow AI music to win a Grammy, provided that the music had a helping human hand. But there’s still confusion over what constitutes AI music.

Media studies professor Jack Hamilton expressed doubts that AI music will replace music made by people. (Photos by Dan Addison, University Communications)

“Most people, including myself, don’t really know what AI music is,” said Jack Hamilton, associate professor of media studies and a pop critic for Slate. He pointed out that “Heart On My Sleeve” still had a human composer. AI just realized the song.

It’s unlikely that AI music will replace big stars. Hamilton doesn’t think there is much of an audience for songs not written and performed by the artists.

“If you’re a Drake fan … it’s not just because you like the way Drake’s voice sounds,” Hamilton said. “You’re a fan of the persona.”

It’s a different story for artists who aren’t already household names.

Darden School professor Mike Lenox warned that AI music’s growing use puts lesser-known musicians at risk. (Photos by Dan Addison, University Communications)

“Non-stars are right to worry,” Mike Lenox, another Darden School professor and a senior faculty fellow at the Miller Center, said.

A lot of these unfamous musicians aren’t making songs that people choose to listen to. Instead, they’re composing background music for commercials, or the song that plays while you’re on hold with the phone company. It would be easy, Lenox said, for companies to just have AI compose those songs.

Lenox and Grushka-Cockayne agreed with Hamilton that there is no replacing megastars like Beyonce and Taylor Swift – at least while those musicians are still alive. An AI-generated recording of the late rapper Biggie Smalls rapping Nas’ “N.Y. State of Mind” landed on social media to rave reviews, though some fans raised ethical questions about using his voice after his death.

Artists are divided, too. The electronic musician Grimes tweeted that she would split the profits with anyone who made an AI-generated song using her voice, though she later retracted that offer. Nick Cave, meanwhile, called AI music “a grotesque mockery of what it means to be human.”

“There are always musicians who have been really interested in experimenting with cutting edge technology. Those musicians tend to be marginal, though,” Hamilton said. Most people making music, he said, are in it for emotional and artistic reasons that seem antithetical to the use of AI.

There are questions about whether the AI music craze will last. For a few years, “immersive art experiences” that used AI to expand existing works by visual artists like Vincent Van Gogh were cropping up in cities across North America. This summer, Lighthouse Immersive Inc., one of the companies behind such experiences, filed for bankruptcy.

Similarly, some of the interest in AI music might be for its novelty.

“Most AI music that I've encountered, personally, is gimmickry,” Hamilton said.

There’s so much hype around AI that companies are going to rush into the technology without having a clue what they’re doing or how they’re going to use it to create value, according to Lenox.

“You’re in a world where every company feels the need, on earnings calls, to mention generative AI and what they’re doing about it, and a lot of it is BS,” Lenox said.

Grushka-Cockayne predicts a day where AI music will become part of our daily lives, like spellcheck or predictive text.

“I don’t think it will die down. The drama around it may just stabilize a little bit,” Grushka-Cockayne said.