First, there was fake Drake. Now, counterfeit Kanye and bogus Bad Bunnys are all over the internet. It seems that artificial intelligence-generated music has arrived.
Some examples are obvious forgeries, like former President Barack Obama performing “Let It Go” from Disney’s “Frozen.” Others, like the Fake Drake song “Heart on My Sleeve” that went viral last month, are pretty convincing.
Marketplace’s Meghan McCarty Carino spoke with Dan Runcie, founder of the media research firm Trapital, about AI’s latest hit and how far this technology has come.
The following is an edited transcript of their conversation.
Dan Runcie: These tools have gotten a lot better in a short amount of time, but I still think we may be a ways away from a song capturing a hit like Drake’s song like “God’s Plan” or song by The Weeknd like “The Hills.” But music has always been a canary in the coal mine for innovation and technology and where things are going. This proves that will also be the case with AI.
Meghan McCarty Carino: What do we know about how “Heart on My Sleeve” was made?
Runcie: This song was created by a TikTok user called ghostwriter977. They were the ones that played it up. They used a combination of available tools to both get the lyrical cadence of a particular artist, but also their voice as well. There are definitely similarities between this song and the type of song that you may have heard on one of Drake’s more recent albums.
McCarty Carino: What makes this song really sound like Drake?
Runcie: With this technology, you have the ability to copy and replicate the singer’s voice cadence. With an artist like Drake, you have enough imprints of what his voice sounds like and what his voice inflection sounds like. You can look at those all as data points, and there are enough of those data points to be able to then create an output that can mirror that with a wide range of words. If you’re looking at the catalog of any song that Drake has done, you could probably find a pretty wide range of words that can give you an idea of how he says different words, how he uses inflection, types of similar lines and anecdotes.
McCarty Carino: Are there any tells in this song that make it clear that this was machine made?
Runcie: There’s still a bit of a gap between this song and a song that Drake could put out that would stay on the top of the hit list, even though this song is derived from the types of things that Drake has talked about, like a lyric of the song that says, “heart on my sleeve, knife in my back.” Sure, that’s similar to how Drake raps about things, but it’s not personal in the same way as his other songs, where Drake may be talking about his dad in Memphis or some of that relationship, like he does in his song “Worst Behavior.” So there’s still some differences, but that said, there are plenty of songs that Drake has released that don’t necessarily have that emotional connection. I think that’s true for any artist. So, it could take a very strong discerning ear to be able to determine the difference, especially if you’re not someone that’s listening to Drake on the regular.
McCarty Carino: After this song went viral, Universal Music Group had the song taken down from streaming platforms, saying that it infringed on copyrights. What did you make of that move?
Runcie: It highlights how complex and difficult this landscape is. The music industry does want to figure out how to get AI right. The thing is, a lot of the action that we’ve seen to date can feel similar to things that the music industry did in the past and how they’ve treated past innovation, like dating back 24 years ago to the rise of Napster. So much of that was predicated on this idea of the label’s copyright and their intellectual property, and they want to be able to protect that.
We’re in a similar type of situation now because if you have an AI-derived track from one of the signed artists, who has benefited from that? Or who monetizes that? There’s a lot of unknowns. The thing is, though, we have the technology, and we have enough ways to determine what’s what. I think this is where the music industry could actually use AI to its advantage.
McCarty Carino: In your opinion, what are some of the downsides of the industry taking a hard-line approach on this and just trying to tamp it down?
Runcie: It can limit innovation and can also have the industry repeating some of the challenges that have happened in the past. AI is only a threat if the industry isn’t willing to adapt.
I often look at producers as an example of this. In the late 1990s and early 2000s, you had producers like Timbaland and The Neptunes who had this beachhead on the producer market because they could charge several hundred thousand dollars per beat, they had access to gatekeepers, and they all had great talent. Because of that, they were able to essentially command a high premium and have a distinct sound, but then music piracy came, and it hurt their business because people that were buying their services didn’t have as much money.
Later, when the technology came in and lowered the barriers to entry, it became easier for someone to be able to make a Timbaland-type beat and put it up on YouTube or put it up on BeatStars and sell it for a few hundred or a couple thousand dollars. While that initially was a hit for Timbaland’s business model, he then found a way to use the technology to level him up. Now he’s the co-founder of a beat marketplace and is owning a company that does the thing that could have easily disrupted him.
McCarty Carino: What does it mean to you to have machines potentially generating music? Do you think that it can ever truly compete with the human art form?
Runcie: I think it’s something that’s more likely to augment and help improve the human art form than I think it will replace it. This is something that we’ve seen time and time happen with new technology. Fifteen years ago, when Auto-Tune was blowing up, people thought that this technology would give anyone walking around the ability to make a hit song. Well, it really didn’t.
When T-Pain did the NPR Tiny Desk performance without Auto-Tune, it turned out that he can really sing. Even with Auto-Tune, you still need something behind that to make it work.
I think this is true of so many things. People thought that once you could democratize the access of making beats that you didn’t necessarily need the talent. Well, you still do. The producers that are rising now may not be going about it the same way that Dr. Dre did the 1990s, but they’re leveraging the tools that are available. So I don’t know if I necessarily think that 100% machine-generated music will rise to the top, but I do think that it can help amplify what’s already there in unique ways that could work to the advantage of both the music industry and the artists themselves.
Related links: More insight from Meghan McCarty Carino
Not everyone in the music industry is feeling skittish about AI and IP.
Days after Universal Music Group had the AI Drake and The Weeknd song taken down, the artist Grimes took to Twitter to encourage people to use her voice to make new music.
She said she’s excited to be a “guinea pig” for this type of experimentation and that she’d evenly split the royalties from any songs made an AI version of her voice with the creator, just as she would with any other creator.
In our conversation, Dan Runcie pointed out that Grimes is an independent artist, so by and large, she owns the rights to her music.
She’s also, famously, a former partner of Elon Musk, who co-founded and served on the board of OpenAI, the maker of ChatGPT. Musk also seems to have plans for his own AI company, while also sometimes saying he believes the technology will destroy civilization.
The future of this podcast starts with you.
Every day, the “Marketplace Tech” team demystifies the digital economy with stories that explore more than just Big Tech. We’re committed to covering topics that matter to you and the world around us, diving deep into how technology intersects with climate change, inequity, and disinformation.
As part of a nonprofit newsroom, we’re counting on listeners like you to keep this public service paywall-free and available to all.