Music & AI: Mix, or Match?

The recent release of the song “Heart on My Sleeve,” with a vocal track featuring the AI-cloned “voice” of Drake and The Weeknd, sent a shockwave through the music industry. But artificial intelligence technology has been in use in the music business since before the latest generation of generative AI tools were widely available. We asked Jessica Powell, CEO of 2021 startup Audioshake, which uses AI to isolate and separate stems from recordings, whether fake Drake represents a break from the past, or more of the same.

Jessica Powell

RightsTech: Audioshake has been using artificial intelligence technology to help music rights owners create new licensing and monetization opportunities since before the current generation of generative AI tools were commercially available. Do you envision those new tools also playing a role in Audioshakes’ business?

Jessica Powell: AudioShake’s stem separation is a different kind of AI. But there can certainly be intersections between what we do and what a generative music system does. So we’re already working with some companies in this space, who are working in partnership with labels and artists to create authorized generative music.

RT: You recently likened the release of the AI-enabled track “Heart on My Sleeve” in part to traditional mashups, which were also controversial when they first appeared. What are the similarities? What are the differences?

View Post

JP: There are many parallels, as I wrote in my Substack post. One key example: The main driver behind an AI cover, as with sampling and remixing, is likely someone paying homage to an artist they love–not trying to rip them off. But I do think there’s an extra nuance here that we should keep in mind: this is someone’s likeness. You’re putting words in someone’s mouth, and those words could be terrible. While a sample can be taken out of context, or re-contextualized, I do think that someone’s voice, appropriated, can’t be thought of exactly the same way.

RT: You speculated that “Heart on My Sleeve” was cobbled together using several different AI technologies, What is the significance of that to how we should regard the track?

JP: On some level, I don’t think it matters much whether it was 100% AI or if a human got involved. These things will get easier and easier to make, so if we’re going to try to take a step back and look at the longer-term implications, we might as well think of it as a full-AI track.

That said, it’s worth pointing out that the majority of AI covers don’t sound that great right now. It’s not surprising that the AI cover that likely had human involvement was the one to break through and go viral. That’s relevant, because as with remixes, the majority of covers are not going to sound great, and are going to get a few streams. The ones that get attention could be dealt with the same way remixes are–with claims or takedowns. Mind you, that’s not to say that the way the remix ecosystem works is ideal–it’s just to say that we already have a framework through which we can consider how to think about AI remixes.

RT: How does the recording industry need to do to prepare for the predicted onslaught of AI-enabled and generated music?

JP: I think the best thing the industry can do is think in specific terms. There is a world of difference between a songwriter using an AI tool to get past writer’s block when they are writing a song, a streaming platform using AI to curate a playlist, and a fan deciding to use another artist’s voice on their track. All use AI, but not the same AI, not necessarily the same training data, or the same intent, use, or distribution. Right now, I do see some conflation happening in the industry as people talk about AI as if it’s one thing. That’s like talking about Google Maps as if it was the same thing as Quickbooks or Microsoft.

I think there are very legitimate questions to be asked about technology and its use. By being specific, we can quickly get to the heart of whether there is a debate to be had, or if this is just another technology (that happens to be based on AI), and that enables a bunch of use cases that we’re comfortable with.

Get the latest RightsTech news and analysis delivered directly in your inbox every week
We respect your privacy.