'You can try to fight it or adapt': How Canadian musicians are responding to AI's rapid advancements

I can't feel the breeze,
Baby, I am a machine,
I don't know what love means.

In May, Australian-born, Los Angeles-based DJ and producer Kito released "Cold Touch," a two-minute electronic track that packs a melodic punch over a propulsive metallic beat. To the untrained ear, the airy, ethereal vocals that lay atop the track sound like Grimes, the Canadian pop star who hasn't released an album since 2020's Miss Anthropocene. While Grimes has released a few original singles over the last three years, you won't find "Cold Touch" on her Spotify profile: the Grimes you hear on Kito's song is an artificial intelligence approximation of the singer's voice, which she made available for anyone to use and create music on via her own AI platform, Elf.Tech. 

In an interview with the New York Times a month after Elf.Tech's April 2023 launch, Grimes explained the intention behind making her voice accessible for others to experiment with: "Feeling really amazing from making beautiful art is something that has typically been behind a gate for a lot of people — extreme amounts of time and energy, years of technical training. I think it's valuable that there's a tool with which, if you have a beautiful idea, you can make a beautiful thing and access that." It's a statement that echoes technological advancements before it, for example when GarageBand launched in 2004 and gave amateur musicians access to easy-to-use music-making software.  

Since Grimes launched Elf.Tech, GrimesAI's official Spotify page, created to house songs that don't feature the real Grimes, has 69 original songs on it, including "Cold Touch," its first and most popular track with more than 700,000 streams. Half of the revenue from those streams have been promised to the artists who created the tracks, as Grimes originally proposed in a tweet on April 23.

A day later, Toronto rapper Killy copied Grimes's tweet verbatim: "I'll split 50% royalties on any successful AI generated song that uses my voice. Same deal as I would with any artist i collab with. Feel free to use my voice without penalty. I have no label and no legal bindings."

"I think AI is about to unlock a whole other world when it comes to music," Killy explained to CBC Music. "My voice model is something me and my team are developing right now ... I see [AI] adding more ammo to creators' tool kits. We're never scared of technological advances; you can try to fight it or adapt." Killy added that some of his fans have already built their own voice models of him, noting, "I've been listening to everything popping up online, some of it is crazy," including a fan-made cover of the Weeknd's "Heartless" using Killy's vocals.   

Grimes and Killy aren't the only artists who have leaned into the future of music and AI's role in it — though they're not close to being the first. AI pioneer Holly Herndon has offered up her voice for cloning since 2021 — but it's not as easy as Grimes or Killy puts it and it's not without legal and moral consequences. Mind you, Grimes's utopian vision is "killing copyright." 

Edmonton-born rapper Cadence Weapon (real name Rollie Pemberton) says he's not surprised to see Grimes's enthusiasm when it comes to using AI to democratize music-making. "[Grimes] has always wanted to be a machine," he said, having briefly crossed paths with her in the Montreal music scene, where she kickstarted her music career. (Just look at the name of Grimes's latest single: "I Wanna be Software.") 

"It's cool that she's put it out there," he continued. But he doesn't agree with Grimes when it comes to her views on copyright: "I think it's a really significant thing, especially for Black artists and people of colour, who have been historically exploited. When we actually have ownership over our likeness, our voices and our lyrics, it's a really important thing." 

Reckoning with AI and appropriation 

Perhaps the biggest example so far of what Pemberton is talking about is "Heart on My Sleeve," a track released on April 4 that was created by a songwriter named ghostwriter977 using the AI-generated voices of Toronto superstars Drake and the Weeknd. The latest landmark release in the history of AI — which stretches back longer than most may realize, especially if we look at the lineage of AI-assisted technology such as Auto-Tune, which launched in 1997 — "Heart on My Sleeve" was posted on TikTok and became an overnight success, earning half-a-million streams before Universal Music promptly took it down. (Universal Music's claim had nothing to do with ghostwriter977's use of Drake or the Weeknd's vocals, though, but instead a copyrighted Metro Boomin producer tag that can be heard at the beginning of the track. Using the likeness of someone's voice remains a grey area that the label couldn't have argued as strongly.) 

"Heart on My Sleeve" and its success illustrated a larger trend that is growing: people cloning predominantly Black rap artists' voices. Many headlines earlier this year focused on the use of AI to create faux covers, for example Kanye West singing a Justin Bieber song, Jay Z performing a Dr. Dre hit, and most unsettlingly, resurrecting Notorious B.I.G.'s voice to cover artists like Nas, Jay Z and more. A debate emerged over whether this would be considered Blackface, a continuation of a conversation that was sparked last year when Capitol Records signed and subsequently dropped an AI-controlled rapper named FN Meka. 

"We're seeing a lot of white people online and their first idea with AI is, 'How can I exploit some Black artists?'" Pemberton explained. "I think we need to reckon with that a little bit more, like why is that? Why is it always Kanye or Drake's voice, or some Black artists that you want to turn into your puppet?" 

There's really no replacing the human experience in art. I feel like what makes my music significant is the human aspect of it.- Cadence Weapon

Issues of appropriation are just the tip of the iceberg that AI in music threatens to expose. On a commercial level, this raises questions and concerns over copyright, legal use of the likeness of an artist's voice, and who is able to monetize these works. On a creative level, it breaks open a conversation surrounding the intention behind someone using this technology and, most importantly, the consent from those whose voices are being used. Even before "Heart on My Sleeve" surfaced, Drake expressed his weariness toward this trend, writing in an Instagram story, "This is the final straw AI," after someone used an AI-generated version of his voice to cover Bronx rapper Ice Spice's hit "Munch."  

'This five-alarm fire around our jobs literally being replaced feels a little extreme'

A common misconception around the use of AI in music is that it will automate music-making to the extent that human beings will be wiped from the equation. But on the contrary, a human operates the AI program, someone inputs the data that helps the AI replicate an artist's voice, and a person writes, produces and records many of the songs that feature an AI vocalist. In response to "Cold Touch," Grimes herself pushed back against critics, tweeting: "See a lot of ppl angry about 'ai music' but kito actually wrote and produced this whole song — the ai part is the sound of the vocal! Hopefully this clarifies how much work she did etc!" 

As an artist, Pemberton says he doesn't feel particularly threatened by AI. "I feel like it's still quite limited," he explained. "And there's really no replacing the human experience in art. I feel like what makes my music significant is the human aspect of it." He notes that AI can't access the emotionality of an artist's work, and it can only learn from the past work of an artist. Part of why he wasn't impressed by "Heart on My Sleeve" when it came out was because it utilized "what Drake rapped like five, six years ago" and "the Weeknd's singing style from 10 years ago."

"As an artist, we grow and change," Pemberton added. "I'm not the same person. It would be weird if I was just doing the exact same thing over and over again." 

But as Pemberton and a number of other artists interviewed for this piece all agree, consent is key. Herndon, Grimes and Killy are all examples of artists who have given the public permission to use their voices to create art, though even Grimes has said that she would take down any offensive art using her voice. An example of this would be a cover of Beyoncé's "America Has a Problem" featuring the AI-generated voice of Ariana Grande, where she sings a racial slur.

While it may be difficult to implement this rule for all artists, especially with so many musicians and non-musicians alike taking liberties with AI voice cloning already — AI-generated music has more than one billion views on TikTok alone — we are gradually seeing record labels reckon with the legalities surrounding the rise of AI. A recent Rolling Stone feature notes that Universal Music Group "envision[s] a system of monitoring and monetization, ideally with the ability for artists to opt out." On Aug. 9, it was reported that Universal Music and Google had begun negotiations over a potential partnership where they'd license out music from their catalogue for AI purposes with compensation going to copyright owners. 

Back in June, Sony Music even created a position for an executive vice president of AI to help coordinate "its business efforts surrounding artificial intelligence," according to a Billboard report. While the position, which is now held by former British Phonographic Industry (BPI) CEO Geoff Taylor, is vague in its description, it's clear that even labels have come to the conclusion that AI is here to stay.

Likewise, creatives working behind the scenes believe that AI is not here to replace them, but to help evolve their roles. This includes the work of producers, engineers and mixers whose work is becoming more and more automated nowadays thanks to various AI-assisted programs.

"I feel like this five-alarm fire around our jobs literally being replaced feels a little extreme," said Philippe Aubin-Dionne, better known as artist/producer Jacques Greene.

"I mean, it's kind of terrifying how much a lot of Top 40 music basically feels like that's how it's made today," he said, noting a shift in technology and music-making that happened well before AI's mainstream emergence this year. Things like sound packs and presets — pre-made building blocks people can use — already enable people to "make 20 tracks a day," he added. 

Many producers have taken an interest in AI, and are discovering ways to incorporate its usage into their day-to-day work. Juno-nominated duo Banx & Ranx — the producers behind rising stars Preston Pablo and Rêve — believe that voice-cloning technology can be used in the demo track stage, to help them imagine what a song could sound like with their acts singing on it. 

"People will be sending us demos for artists sometimes," Yannick Rastogi, a.k.a. KNY Factory explained, "so we can take someone's a cappella, put it in AI, and have a Preston Pablo singing that record." Rastogi said he wouldn't go as far as replacing his artists with AI-generated versions of their own voices — he likens that to buying a counterfeit watch — but argues that it's a tool he should learn in order to keep up with the times. "We don't want to become dinosaurs ... we just have to make sure we don't lose the humanity in our creations." 

"That's the kind of stuff that makes me feel OK and excited about it all," Greene added, "that some people are really thinking creatively and with a critique about it, and wanting to be proactive because Pandora's box doesn't close up again. We now live in a world with [AI]." 

It's definitely a dangerous inflection point right now where I think music, TV and journalism have all been devalued.- Jacques Greene

Rastogi's creative partner, Zacharie Raymond, professionally known as Soké, sounded both inspired and overwhelmed when he remarked that "it just feels like technology's going really, really fast." 

We're now witnessing the music industry's adaptation to AI advancements in real time. On June 16, the Recording Academy announced a new rule for the Grammy Awards that stated "only human creators are eligible to be submitted for consideration for, nominated for, or win a Grammy Award. A work that contains no human authorship is not eligible in any categories." Less than a month later, on July 4, Recording Academy CEO and president Harvey Mason Jr. clarified that rule in an interview with the Associated Press: "Here's the super easy, headline statement: AI, or music that contains AI-created elements is absolutely eligible for entry and for consideration for Grammy nomination. Period. What's not going to happen is we are not going to give a Grammy or Grammy nomination to the AI portion." 

And now TikTok, currently the No. 1 place for music discovery online, has revealed its own AI music generator called Ripple. With more than a billion users on its platform, the app is now rolling out a beta program that allows people to transform humming and singing into a complete song with full instrumental. Almost 20 years after GarageBand's arrival, Ripple, Elf.Tech and more programs coming out of AI's recent boom are promising ease and access to music creation for more people than ever. 

With approximately 49,000 new songs uploaded to Spotify alone on a daily basis, it's hard to tell how much more musical saturation we can handle — but it doesn't appear that we'll be slowing down anytime soon.

"It's definitely a dangerous inflection point right now where I think music, TV and journalism have all been devalued," Aubin-Dionne warned, noting Facebook's role in "decimating journalism websites" and the ongoing Hollywood strikes. "It kind of feels, without wanting to be too cynical, like it's purposefully done by tech companies ... we're showing that there's easier ways to create something that's kind of mediocre, but you can definitely make more of it and many people might just settle for it."

"I don't want to watch that happen," he added. "I choose to believe that enough people are discerning and curious enough to not fall for it."