AI Music Is On The Charts. Where Does It Go From Here?

AI-generated songs are breaking onto the charts, and music labels are pivoting from lawsuits to partnerships with AI startups. What comes next?

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  WNYC Studios is supported by Wise. Wise, the smart way to manage the currencies you need around the world. With a Wise account, you can send, spend and receive in up to 40 currencies with only a few simple taps. Be smart, get wise. Download the Wise app today. Terms and conditions apply.

• 0:17Ira Flatow?

  Science Friday is supported by MathWorks, creator of MATLAB and Simulink software for technical computing and model-based design. MathWorks, accelerating the pace of discovery in engineering and science. Learn more at mathworks.com.

• 0:33Speaker 00

  WNYC is supported by Bear. Science is about asking questions, testing, trying again. It's not easy, but it's relentless. It evolves, and that's why science delivers. Learn more at sciencedelivers.com.

• 0:51Flora Lichtman?

  I'm Flora Lichtman, and you're listening to Science Friday. In recent months, AI music seems to have exited the realm of novelty act and moved into the world of having living, breathing fans. But what's the impact gonna be? Science Friday producer and musician Dee Peterschmidt is here to investigate. Hey, Dee.

• 1:11Speaker 03

  Hey, Flora. AI music got on my radar last year because I kept getting these videos on my algorithm.

• 1:17Speaker 04

  Inventor of the road roller was a genius. He made a 20-ton machine create 60 tons of force. The secret is that the roller drum isn't a solid piece of metal.

• 1:28Speaker 03

  AI-generated songs describing how various pieces of heavy machinery work. Pretty mainstream stuff. But for the first time last year, some AI-generated songs actually got onto the charts, like this one from Zainaya Monet.

• 1:49Speaker 03

  And some pretty big names in the music industry have gotten pretty vocal about using it, too, like Timbaland.

• 1:54Speaker 05

  It's like an assistant. When I do a beat, and I'm like, yo, how would you take these drums and rearrange it this way? And I'm like, oh, I would have never heard it that way.

• 2:03Speaker 03

  So what's going on right now at AI music companies like Suno? And is this just another tech upgrade to the music-making process? Or is it something else? I wanted to call up one of the journalists I follow on this topic, Kristen Robinson, senior writer at Billboard, who covers AI in the music industry. Hey, Kristen.

• 2:21Speaker 06

  Hey, Dee.

• 2:22Speaker 03

  I was wondering if you also had a moment last year where you were like, oh, this stuff has kind of gotten to another level now.

• 2:29Speaker 06

  You know, I think it was around Zainaya Monet, who you mentioned in the intro. I think Zainaya Monet was a real turning point. But you could you could point to a few different turning points. It just kind of felt like a lot of stuff started happening really fast. So to back up a little bit before Zainaya Monet, you know, June hits and I find that there's this song on TikTok called "A Million Colors" by Beanie Prey.

• 3:00Speaker 06

  I was seeing it in TikTok clips with Kylie Jenner doing her makeup to the song. And I realized that the song sounded kind of weird. And it turned out it was an AI song. And it was towards the top of the viral chart on TikTok. And then that just kind of felt like the first domino. And then things just totally got out of hand. Later that summer, this band called The Velvet Sundown, which was fully AI-generated music and also AI-generated images to correspond with it, really caught fire online. And then Zainaya Monet, it was in September. She became a big headline for us at Billboard because she signed a reportedly multimillion-dollar record deal

• 3:44Speaker 06

  with a traditional music company called Hallwood Media, who's known for just working with regular artists previous to this point. And I think that a lot of people in the music industry considered Zainaya Monet's signing and the fact that her songs were starting to climb on our gospel charts as a really big turning point when AI music has suddenly arrived.

• 4:04Speaker 03

  When we say she signed with this label, who is the person getting the money here?

• 4:10Speaker 06

  OK, that's a great question. So Zainaya Monet is the AI-generated avatar and character created by a woman named Talisha Nikki Jones. Talisha lives in Mississippi. She considers herself to be a poet, but isn't someone who knows really how to get to a finished product of a final song. And so what they say Zainaya Monet is is like a character for her to express her poetry through song. And so the person signing that deal would be Talisha Nikki Jones. And so the royalties would go back to her. And she has a manager as well. And what they would probably say is that these AI-generated

• 4:54Speaker 06

  characters or personas are no different from like how Damon Albarn created the Gorillaz with their little cartoon characters that kind of represented the band. They think that this is a way for artists to maybe express themselves in genres that don't typically follow what they're known for. So they would probably say that this leads to more experimentation. And yeah, it's very interesting. There is a woman behind Zainaya Monet.

• 5:25Speaker 03

  Yeah, I mean, can you talk a bit more about what kind of genres we're seeing AI music kind of glom onto? We got gospel. We've got these like weird, heavy machinery videos. What other genres are coming up?

• 5:42Speaker 06

  I am seeing a lot in the gospel Christian realm. I'm seeing country music. That A Million Colors song that I mentioned is more of like kind of a doo-wop, throwback 50s rock song. I think what I'm really seeing is that it's going for niche genres that tend to be fairly formulaic. Country music is not a super complex genre. Of course, it has so much heart to it, and that's why we all love it. But the chord structure is usually pretty simple. It's usually a verse-chorus, first-chorus, bridge-chorus kind of structure. There's not a ton of experimentation going on in that genre, and the lyrics tend to follow specific tropes. And I think that makes it a little bit easier to make a realistic sounding.

• 6:29Speaker 06

  AI song in those genres. These are, the French streaming service has done a lot of research in this field and they've, they've said publicly that their research shows that 97% of listeners cannot tell the difference between an AI-generated song and a human-made song. So I think it's very possible that some of these AI songs are being listened to and consumed by people who are not fully aware that they're listening to AI music.

• 6:53Speaker 03

  Well, I mean, do you think AI music has gotten like quote unquote good now? Can you, are you one of those 97% people who has trouble telling the difference?

• 7:03Speaker 06

  Sometimes it is hard to tell. I think the big tell still is that the audio quality isn't fully, like, I don't even know how to describe it. It's a little bit of a scratchiness or-

• 7:13Speaker 03

  It's a little digitally sounding.

• 7:15Speaker 06

  Yeah, yeah, yeah, yeah. It's like pixely, I don't even know how to describe that.

• 7:18Speaker 03

  Yeah, the audio version is pixelated.

• 7:20Speaker 06

  You know what I mean, yeah. And I, I think that that's the big tell. And if you're in an environment where you don't have good headphones, like if you're listening on your iPhone speaker, I think it's actually pretty easy to get fooled now. I mean, I guess I would consider myself part of the 97%, although I think I can discern a lot better than your average person, just based on the nature of my job.

• 7:42Speaker 03

  Yes. You've talked to musicians like Imogen Heap, Charlie Puth, about their use of AI. What sense do you get for musicians about how they feel about AI music? And maybe we can start with those examples first.

• 7:54Speaker 06

  Well, Imogen Heap has always been on the cutting edge. If anyone who's listening here is familiar with her work, she's always been both a musician and a technologist. She really feels that technology can make her art more impactful and take her to new places creatively. So she is leaning in pretty hard, but she is still very concerned about models that train AI music models on works like hers without any compensation for those who they're training on. So she still tries to stay away from companies like Suno, which currently have models that are being trained on copyrighted material without licensing or compensation for rights holders.

• 8:39Speaker 06

  But yeah, I'm, I'm seeing musicians really divided. I don't really think you can say- Like, you know, everyone's doing this or everyone's doing that. I would say that a shocking number of professional songwriters and producers have been telling me, mostly off the record, that they are using Suno as part of professional songwriting sessions now. And so a lot of them have posited to me that there are probably songs on the Hot 100 right now that have bits and pieces of AI-generated material that is not disclosed.

• 9:14Speaker 03

  Right.

• 9:15Speaker 06

  A little crazy to think about that.

• 9:17Speaker 03

  I want to go to Suno. And can you give us an idea of who the main AI music companies are? And, you know, we've talked about these meme songs. We've talked about it helping with the production process. What exactly are they selling to people?

• 9:32Speaker 06

  Yeah, so when we think of like, you know, generating songs at the click of a button, that is really dominated at this point by Suno. It's an AI music startup. Suno is quite controversial in the music industry because people feel very threatened by them. I obtained an investor pitch deck of theirs back in November and reported that seven million songs are being generated on Suno every day. That kind of scale scares musicians quite a bit, just in terms of, although those seven million aren't necessarily making it onto streaming services, a lot of them are, like, some of them are. And that crowd, potentially crowds out works made by human musicians. So Suno is a big one. Udio is another big one.

• 10:17Speaker 06

  They did kind of the same thing. You can type something into a text box and then out pops a song. Udio is now pivoting to do AI-powered remixing of already-made songs. This is a very popular category in AI music right now. Spotify is even getting into this realm soon. And basically, what this means is that with licenses in place, you would be able to take two of your favorite songs and create mashups. Maybe you remove the vocals so you can do a karaoke version. And you can speed it up. You can slow it down, all these kinds of things. So you can play with music that already exists.

• 10:52Speaker 03

  Well, it's funny because last year...

• 10:57Speaker 03

  The main, the major music labels were trying to sue the heck out of these AI music companies.

• 11:02Speaker 06

  Yes.

• 11:03Speaker 03

  Now they're partnering with them. What happened there?

• 11:05Speaker 06

  Yeah, so I think the music companies are really realizing that they can't make this go away, and so they need to find a way to extract value from it. I think another thing to keep in mind is that very recently, like within the last decade, two of the three major music companies became publicly traded companies. So they're probably getting a lot of shareholder pressure to innovate, to integrate AI and to capture value there. They don't want to be seen as weak. They don't want to be seen like they're behind the ball. So I think that that's also, you know, one of the reasons why they have been so willing to try to find reconciliation.

• 11:43Speaker 03

  Right. The music industry has gotten left behind a lot in the past in regards to tech and seems like they're changing their tune. So with all these recent deals, do you have a sense of where this is all heading? What do you have your eye on this year?

• 11:58Speaker 06

  Interestingly, the music AI game has mostly been dominated by startups. My take on that situation is that I think that music is a very hard thing to generate, and it's also not something that's a huge moneymaker. So I think it's been largely ignored by your Open AIs and Googles of the world until now. But Google has launched Lyria 3, its latest AI music model on Gemini. Still not as good as Suno or Udeo, but who knows in the next year how it will develop. And they also acquired an AI music company called Producer AI. So I have my eye on Google for sure, and I also have my eye on these new models from Suno and Udeo.

• 12:39Speaker 03

  Okay, we'll keep an eye on them, too. Kristin Robinson is the senior writer at Billboard who covers AI in the music industry. Thanks, Kristin.

• 12:46Speaker 06

  Thank you.

• 12:47Speaker 03

  Okay, stay with us because after the break, we have one of the first musicians who experimented with algorithmically generated music back in the 70s, and we'll hear her take on AI music.

• 13:02Speaker 00

  Science Friday is supported by PBS, presenting Going Wild. If you love learning about the natural world and meeting the scientists who help us understand it, be sure to check out the new season of Going Wild with Dr. Ray Wynn Grant. It's the chart-topping and Webby Award-winning podcast from the folks at Nature on PBS, now with a video companion series. In this season of Going Wild, host acclaimed wildlife ecologist Dr. Ray Wynn Grant heads out into the field to explore the animals, people, and stories that make nature worth saving. Stops include a hibernating black bear den in Arkansas and an underwater shipwreck site off the Florida coast. If you're looking for a place to start, check out the season's debut episode, where Ray joins beloved biologist and influencer Jordan King in the Florida Everglades to learn about

• 13:47Speaker 00

  his unlikely journey into science and how to study a wetland when it's not very, well, wet. So don't miss out. Follow Going Wild with Dr. Ray Wynn Grant on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts.

• 14:00Speaker 07

  And from the Alfred P. Sloan Foundation, working to enhance public understanding of science, technology, and economics in the modern world.

• 14:13Speaker 03

  And now, a person with a one-of-a-kind perspective on AI music, musician Laurie Spiegel, a pioneer of electronic music and also of algorithmically-generated music. And like today, it raised some eyebrows at the time. She wrote code for some of the first computer music technologies, and her 1980 album, The Expanding Universe, is considered one of the greatest ambient music albums of all time.

• 14:43Speaker 03

  Another song from that album is even on the Voyager spacecraft's golden record. Laurie, it's so great to have you here.

• 14:50Speaker 04

  Hi, I'm glad to be here.

• 14:53Speaker 03

  Did people think what you were making in the 70s and 80s was music? Kind of like this AI conversation right now? Did you get shade when you got into this?

• 15:05Speaker 04

  There was a lot of heavy anti-computer sentiment back then because computers belonged to the most oppressive of organizations only. They weren't personal computers yet. And it was the government, the banks, the insurance companies, the military who had computers. And the computers, innocent things that they were inherited, the image of the oppressiveness of their controllers in the public eye. Because computers, you know, it was, they were cold, they were inhuman, they were hostile to the arts. They were not the warm cuddly little laptops that we are used to at this point.

• 15:48Speaker 04

  So that I was often accused of dehumanizing music. But of course, technology is the most human thing around. I mean, we are by far the animal that does the most technology.

• 16:04Speaker 03

  You know, I think in the late 70s, you worked on an algorithm to kind of replicate Bach's harmonic style.

• 16:12Speaker 04

  Yeah, Bach is just a superlative ideal for me and inspiration. And so I studied the harmonic progressions of the Bach chorales extensively and wrote a way simplified compared to the mind of Bach algorithm that basically generated harmonic progressions that I felt were meaningful.

• 16:45Speaker 03

  Yeah, and obviously, he's a super mathematic kind of composer. And it makes some sense to be like, how can I? Translate this into an algorithm. I mean, the other, the other side of the, the modern AI music side is, you know, there have been, there have been these studies of people who use these large language models, experiencing something called de-skilling, where, you know, you end up starting to rely so much on these models that you kind of end up outsourcing a lot of your own skill to them, and then that skill atrophies over time.

• 17:18Speaker 04

  And yet, different skills are evolved during that process because the, the writing of the prompts for an AI system is in itself at the very first stage of becoming an art form, I think. But it's quite different from the moment to moment generating of sound in response to your momentary emotions, the, the self-expressiveness of playing, playing music. And that is something that, that the way AI has been doing by giving a prompt and then waiting for fabricated results is quite different from, I mean, the, the expressive nature

• 18:00Speaker 04

  of playing an instrument is, is, it's visceral, it's tactile.

• 18:05Speaker 03

  I mean, I've heard some music producers, you know, have talked about using Suno, one of these products, and AI music because they, they don't want to be left behind. And I've seen that language being used with other AI tools. What do you make of that in terms of, I mean, did you, did you feel like you were going to get left behind, you know, back in the seventies if you didn't engage with, you know, computer programming and making fresh music?

• 18:33Speaker 04

  Oh, just the opposite. I was, I was kind of way out ahead to the point where it was impossible to explain to people what I was doing. I was not at all left behind. I was like on the lunatic fringe. I couldn't explain to people. People would say, oh, you do music. What kind of music do you do? And I would, I would say, well, I'm using computers. And they would immediately, like, their expression would change and they'd want to change the subject too. In the arts, it's not a matter of keeping up. It's a matter of something honest and authentic coming from inside of you that you can embody in an experience external to you that you can share with other people.

• 19:16Speaker 04

  Everybody's always trying to like keep up with what is new. That's not what makes high-quality artistic expression. It has to be from inside of you. The music itself is what's important. And that's not something which is reliant on any individual technology. It's gone through many centuries of evolution of different technologies. And it's still obvious to us what's really good music from the Renaissance that moves us and grabs us, or the early 20th century, or whatever. It's what it does for us, you know, what music does for us that's important. So I don't think it's really worth worrying about keeping up with tech.

• 20:01Speaker 04

  It's just how you use it.

• 20:04Speaker 03

  Yeah, well, it seems like so many things with AI is forcing us to ask like these really basic questions about the things that we like and why exactly we like them. I know you were just talking around that, but I mean, what does, what does music mean to you?

• 20:23Speaker 04

  God, I don't know. I mean, the question that I posted at the beginning, the question of what the purpose of music is for people and what parts of what music does to us are these AIs able to satisfy. They obviously can generate music-like material on demand, but it's not necessarily the expression of emotion or feeling. I really do want to play with them a bit more. I know that the writing of prompts is a very indirect way of making music, much like

• 21:05Speaker 04

  writing all these little dots on staff paper with a pencil. And then you get back a result which is not what you had anticipated because they're not really interactive. Emotions are kind of a... We don't understand them very well yet, but they are rock bottom an essential component of music. And this is where the AIs kind of fall down. They don't have them.

• 21:34Speaker 04

  And while they will probably figure out how to trigger them and evoke them eventually, and really good prompt writers might be able to do that. It still is very much in its infancy, these non-interactive generative parrots, I guess you could call them.

• 21:57Speaker 04

  They speak the language that they have read all over the net or throughout the repertoire, and they parrot it back. But they don't understand it on a gut level that we humans experience it.

• 22:11Speaker 03

  Laurie Spiegel, a pioneer of electronic music and algorithmically generated music. Thanks for being with me, Laurie.

• 22:18Speaker 04

  Thank you for having me.

• 22:19Speaker 03

  By the way, one of Laurie's best-known pieces of software, Music Mouse, which she made in 1986, recently got re-released on modern computers. It's like an interactive instrument you play with your mouse, where you basically drag your mouse around a musical grid, and it makes these fun chords and melodies. If you want to try it out, you can find a link to it on her website, sciencefriday.com/music.

• 22:41Flora Lichtman?

  Thank you, Dee. This fantastic episode was produced by Dee Peter Schmidt. And listeners, if you have thoughts or feelings on this or anything else that we cover, you are always here for it. 8774 Sci-Fry. Thank you for listening. We'll see you tomorrow.