It's our first-ever crossover episode! This week on Have a Nice Future, Gideon Lichfield and Lauren Goode sit down with Puja Patel, cohost of The Pitchfork Review podcast, to discuss how AI is changing music. But first, they start with a pop quiz: Can our hosts differentiate between artists and their AI imposters?
Check out some of our stories about the intersection of AI and music, like how musicians are finding ways to harness generative AI creatively, the AI-generated music flooding your streaming platforms, and the bots listening to those AI-made songs.
You can always listen to this week's podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here's how:
If you're on an iPhone or iPad, just tap this link, or open the app called Podcasts and search for Have a Nice Future. If you use Android, you can find us in the Google Podcasts app just by tapping here. You can also download an app like Overcast or Pocket Casts, and search for Have a Nice Future. We’re on Spotify too.
Note: This is an automated transcript, which may contain errors.
Gideon: I sound a bit manic? OK.
Lauren: [Laughter] Hi, I'm Lauren Goode.
Gideon: And I'm Gideon Lichfield. This is Have a Nice Future, a show about how fast everything is changing.
Lauren: Each week we talk to someone with big, audacious ideas about the future, and we ask them: Is this the future we want?
Gideon: This week we're talking about how generative AI is shaking up music with Puja Patel, the editor in chief of the music site Pitchfork.
Puja Patel (audio clip): I think people are getting more and more particular about having very clear, specific tastes, and also about not letting the computer trick them.
Lauren: Gideon, I wanted to start the show off a little bit differently today.
Lauren: It's time for a pop quiz.
Gideon: Oh no.
Lauren: So you are the editor in chief here at WIRED, and you've been talking a lot about AI, so I wanted to see how good you are at telling regular human-made music apart from AI-generated music.
Gideon: I mean, I can barely tell music by one human apart from another sometimes. So, uh, you know, you might be disappointed, but I will do my best.
Lauren: OK, here's the first track.
[Clip of AI Rihanna singing “Cuff It”]
Lauren: Let's hear the second one.
[Clip of Beyoncé singing “Cuff It”]
Lauren: First, I'm curious if you know who the artist is.
Gideon: I have no idea. None whatsoever.
Lauren: Gideon! Not a card-carrying member of the beehive, are you?
Gideon: I am so not.
Lauren: That was Beyoncé.
Lauren: But which one was actually Beyoncé?
Gideon: So was one of those definitely AI and the other one was really Beyoncé?
Gideon: Uh, wow. Um, I'm gonna say the first one was AI, just because I liked the second one better.
[Correct Answer bell rings]
Lauren: That is correct.
Gideon: OK. All right. So taste actually plays a part here.
Lauren: You get to keep your job as the editor in chief of WIRED.
Gideon: Thank God, I was so worried.
Lauren: What about the second one did you like more?
Gideon: Here, I'm gonna have to use like music-y words that I don't know the meaning of, but I just felt like it was, I don't know, the sounds were more balanced somehow. More harmonic, more … I'm making crap up here. Uh, I just like the sounds better. They felt more like a thought had gone into that song.
Lauren: Yep. Fair enough. Let's move on to the second pair, which I think is going to be a little bit more challenging. All right, here's the first track.
[Clip of “Controlla” by Drake]
Lauren: OK, here's the second one.
[Clip of AI Arianna Grande singing “Controlla”]
Gideon: I'm gonna say that the female voice is the AI-generated one, just because it sounded a little more Auto-Tuned to me. But honestly, if you'd played both of those songs somewhere and I didn't know that one of them had to be AI generated, I would've just said they were both human.
[Correct Answer bell rings]
Lauren: You are correct. The female voice is AI-generated. The first one was real. That was actually Drake.
Gideon: OK. So I may be completely useless at music, but at least I can tell an AI from a human when they're singing. That's something.
Gideon: That's what the editor of WIRED should be able to do. Right?
Lauren: OK, so this next pair is interesting because this person has really been at the forefront of what I'll call the “If you can't beat ‘em, join ’em” movement. So I’m tipping you off a bit.
Gideon: This is Grimes.
Lauren: OK, let’s hear two tracks, and you can determine which one is real and which one is not.
[Clip of “Cold Touch” by GrimesAI]
[Clip of “Laughing and Not Being Normal” by Grimes]
Gideon: This is a complete coin toss for me. I honestly have no idea which one. So you’re saying one of those songs is an actual Grimes song, and the other one is something that somebody made using Grimes’ voice, yes?
Lauren: Yes. It is a tough one.
Gideon: As someone who really isn't familiar with Grimes’ work, I honestly don't know which one of those is which.
Lauren: Yeah. Her sound is typically a very ethereal, kind of manufactured manic pixie sound.
Gideon: The second one felt more “manic pixie” to me.
Lauren: And so does that mean that's really her?
[Correct Answer bell rings]
Lauren: You are correct.
Gideon: Wow. Three out of three?
Lauren: You actually did quite well. Yeah. You were using, like, context clues, and you were right. You probably did great on the SATs. Did they have those in the UK?
Gideon: No, we didn't have SATs. I was good at university, yes.
Lauren: And I could see why. I'm impressed you got them all right.
Gideon: I mean, not thanks to actually knowing anything about music. I have to say, I feel like I have a decent ear because I did play instruments when I was a kid. But, um, it was just as much guesswork as anything else. I wouldn't have been surprised if I got them all wrong.
Lauren: But I think you did a really good job of using context clues to figure out which way was up or which song was real. And honestly, that's my biggest question right now. Like what is real? If something is heavily Auto-Tuned or has a lot of artificial elements in it, does that make it “fake,” or is it only fake when it's being generated as something totally new without any kind of human participation? And like, what does this mean for the future of music?
Gideon: I guess there are degrees of fake, right?
Lauren: Well, all this is really complicated stuff, and that's why I'm very excited about today's guest. We've asked Puja Patel to join us on Have a Nice Future this week. She's the editor in chief of Pitchfork, which like WIRED is a part of Condé Nast.
Gideon: That’s right. And as well as running Pitchfork, she hosts her own weekly podcast, The Pitchfork Review, where she and her colleagues review … music, of course.
Lauren: Now, we brought Puja on the show because a few weeks ago it became really apparent that AI music was taking over our feeds. A new song by Drake and the Weeknd dropped, and everyone was talking about it because it wasn't actually by Drake or the Weeknd.
[Clip of “Heart on My Sleeve”]
Lauren: Someone had used AI to create the song entirely anew. And honestly, it fooled a lot of people. Well, maybe not you, Gideon.
Gideon: I probably wouldn't have even been able to tell you who Drake or the Weeknd was at that point, so, uh, it doesn't matter. Shortly after that, Grimes announced that she'd be allowing her fans to create their own AI-generated songs using her voice, and that she would split any revenue with them 50-50.
Lauren: And even though Gideon and I are not music nerds, it made us think a lot about what our AI music future looks like.
Gideon: I mean, you don't have to be a music nerd to realize there's a lot at stake here for established artists, for up-and-coming ones, for the record labels, for the streaming services, and for the rest of us who just like to listen.
Lauren: Yeah, and since there's so much ground to cover, today's entire episode is going to be a conversation with Puja, and that's coming up right after the break.
Gideon: Puja, thank you for joining us here on Have a Nice Future.
Puja: Thank you for having me. I'm excited to talk AI.
Lauren: Puja. Just to kick us off, what is AI-generated music? How is it different from remixes and stems and AI-enhanced music of the recent past.
Puja: The thing is, it's not so different from a lot of that, because AI is infiltrating every single aspect of music. So when you think about what makes up a song that has AI in it, there are tools that range from like, OpenAI—this is my favorite story actually of recent history. OpenAI slash Jukebox, where there's this, I don't know if you're familiar with David Guetta, who is this like festival, EDM, forward DJ, and there's this video of him giddily explaining how he got Eminem to work with him on this song that he debuted at a festival. And what he did was, he used ChatGPT to write lyrics in the style of Eminem. And then he used a different program to input those lyrics that then were delivered in the sound of Eminem's voice.
[Clip of David Guetta remix]
Puja: And then he debuted that song at a festival, and he was just like, well damn, isn't this brilliant? Um, so that's like the OpenAI version of it. And then there's the kind of more complicated version, which is tied to how that Weeknd/Drake song “Heart on My Sleeve” was released, which is where a person had basically created a library of sounds using Drake's voice and using the Weeknd's voice, and then made his own melody, made his own beat, and then used his library to create a version of what might sound like Drake or might sound like the Weeknd to create an entire new song. So there's just like a million different ways to use this technology. Some feel a little more aggressive than others. Um, some have, I think, higher ethical concerns than others, and I will say this isn't necessarily new. There have been musicians doing this sort of interpolation and manipulation of existing recordings for a very, very long time. I think it is just the fact that this is now so easily accessible to the public that has made it a little more confounding and concerning and strange for everyone.
Gideon: All right, so people are like, seem to be freaking out about the idea that there might be a million songs out there all using Drake's voice and none of them are sung by Drake. But what, if anything, should people be freaked out about here? What do you think are the concerns, and what are not the concerns?
Puja: I mean, I think this is the same concern that has been prevalent in art since the dawn of time. The bigger concern is around ownership. And, respectfully, I don't want to hear 20,000 Drake songs. Like I don't care about that. I don't care about half of Drake's songs that already exist that are made by him. So I don't see it becoming like a value proposition in a way that supports the overall brand of the artist in a financial or recognition way. And I think the big freak-out and the big existential crisis here really is about money and ownership.
Gideon: But like, but for whom? Because, right, as I think you're saying, if a million people make copies of Drake songs, it won't make Drake any richer and it won't make him any poorer. And the vast majority of the songs nobody will want to listen to. And even the good ones, even the good ones, nobody's gonna say, oh, wow, what an amazing original song. Because it will basically be an incarnation of Drake. Like without Drake, that song wouldn't exist. None of what was great about it would exist. So all of the stuff that is copied off Drake or a famous artist like that seems like it's an irrelevance in some sense.
Puja: I mean now we're broaching the subject of copyright, and since the dawn of time, people have been questioning whether a sample is something that is worth being fought over. Part of the reason this is so sticky is because the law is centered to protect humans. When we were reporting about this at Pitchfork, I think a month ago, the copyright office issued some new declaration around AI, and they basically said that something can be protected by copyright if it had immediate human interaction with it, which meant like it was created by a human. And where that then becomes tricky is because if you significantly manipulated AI as a human to create something that sounds like a lot of other things, that still is protected by copyright. But we haven't seen something go to court. I think the bigger conversation around AI is around AI training and copyright. Like, if you as a person are training a computer to be something else, using a source library of an artist, like that is the bigger copyright question right now.
Lauren: Mm-hmm. Because then the computer, or in this case the model, becomes the product of human authorship, whereas the thing itself that's spit out by AI might not be the product of human authorship, which would be protected by copyright law.
Puja: Right. But then also, let's say that you are training the model using Drake's vocals without Drake's consent, um, that becomes the point of contention.
Gideon: Is there anything that you are excited for in the world of AI-generated music? I mean, it sounds like mostly so far we're talking about the musical equivalent of fan fiction.
Gideon: Um, and, which is great. And, you know, the fan fiction universe is a, is a wonderful and vibrant thing. But is there anything for genuine musical creativity, new forms of music, uh, new ways for original artists to emerge that you think AI enables?
Puja: Absolutely. I mean, I think we underestimate how much musicians are using this already. Just from a financial perspective, it opens up doors for people who are under-resourced, who aren't signed to a label, who don't have the financial backing to create something in the way that they're able to. I was reading an interview with a producer once who basically said, “You know, I am a musician searching for a voice. You know, all of my work is behind the scenes. I'm like constantly looking for the perfect singer and vocalist to be in my music. If I can help source my ideal voice, that changes what I'm able to create and how I'm able to think.” And that to me sounds wild, but also I think that just like with the advent of anything, there's lots of opportunity here.
Lauren: What about the artists who are embracing this by saying they'll do a rev share with people who creatively use their voice? Grimes comes to mind.
Lauren: What do you make of that?
Puja: So Grimes is leading the path forward for this. Before we talk Grimes, I do wanna shout out Holly Herndon, who is this academic and electronic musician. She really did this first, or did this early on. But Grimes has basically declared that she will make available her voice to anyone who wants to use it and that music would be able to be uploaded to any streaming platform under the moniker GrimesAI-1, and any royalties that would come from that music, which she would be able to track because of the way that those stems and such are coded, would be split down the middle. This is part of Grimes’ larger philosophy around, you know, the idea that technology and humanism are in conversation, that music is a conversation with audiences, that there should not be copywriting, and like copywriting is a form of gatekeeping. So her, her platform, Elf.Tech—I'm smiling …
Gideon: We can hear the smile. It's an audible smile.
Puja: So her platform, Elf.Tech, basically seeks to allow anyone to use any of her vocals in any of their music. Um, I think it's a commendable act. I again ask who wants to hear 10,000 Grimes songs. But I'm curious to see how that moves forward.
Gideon: I mean, there's no downside for her I think.
Puja: Yeah. If anything it helps her. One hundred percent.
Gideon: Maybe it helps her in terms of exposure. It certainly helps her in terms of, you know, being the first to do this.
Gideon: And as you say, maybe nobody wants to hear 10,000 Grimes songs, but what if some people do and are willing to pay for them, then she wins and so do the creators.
Puja: Well, I think the bigger possible, like lucrative success there is if something goes viral or something becomes a streaming hit or something makes it onto the chart. I think she's opening up business opportunities for herself, because in the chance that one of those 10,000 songs lands, that changes things for her too.
Gideon: Can we talk about the other end of the music industry? I have a cousin who is a musician. He made some of his living from recording, writing music for advertising jingles and things like that. And I think a lot of working musicians are doing that sort of thing. Do you think those livelihoods are threatened by generative AI? Is it, you know, in, in much the way that people are worried that stock photography is going to go the way of the dodo because it's just really easy to generate generic images that can be dropped into, to simple things?
Puja: Absolutely. I mean, I think that's the first to go. Um, again, speculating here, but there was a musician who talked to Pitchfork, who basically said that a lot of these working musicians couldn't make money, especially in the pandemic, from live gigs anymore, from releasing new records. And so much of their work became tied to sync music, exactly like you're saying, Gideon, or writing music for TV, films, commercials, so on and so forth. But the interesting thing that was being talked about then is, uh, how much of that work was expected to be free or at a much lower cost than what writing and selling your own music would get you back. So I believe the platform is called Amper, but there is an AI platform that is specifically making music to soundtrack anything from home videos to commercials to trailers for movies, to anything else, and you are simply manipulating the vibes—the vibes and the instruments there.
Gideon: Do you worry about what that means for people who try to get into music as a profession? Does that close off an avenue that people might have followed in the past, in the hope of achieving stardom?
Puja: Yeah. I think that this is all, this is kind of why the conversation around ownership is looming so large, because I think that—I don't know that people were making very much money from that to begin with.
Lauren: It seems like there's a lot to unpack here about how this affects the artists. Do you have a sense of how this is affecting the streaming platforms?
Puja: I will honestly say that streaming, as always, has no idea what they're doing. [Laughs]
Lauren: Say more about that.
Puja: OK, so maybe you saw this: Boomy, which is one of those AI music generators, has uploaded something like 14 million songs to Spotify or has created about 14 million songs, of which many, many, many have made it onto Spotify, and Spotify recently declared this mission to say we are diligently working toward removing these AI-created songs that have been uploaded, because our solidarity is with human creators. Within the span of several weeks that was backpedaled, and now Spotify and Boomy are in contract negotiations for how to upload AI music into Spotify.
Lauren: I mean, I imagine in the long term, it's actually quite bad for a company like Spotify to have tons of sludge on its platforms. Even if they got a short-term boost in number of tracks or listeners, because the competition between the streaming platforms is so fierce, and if customers are paying $10 a month, they want a high-quality experience, and a lot of people still want to listen to what they might consider traditional music. Right? Like does it—I, I can't see how it benefits Spotify in the long term to just become the platform for AI-generated tracks.
Puja: I mean the same thing happened with YouTube—I think with Jay-Z a couple of years ago when some clever coder figured out how to make Jay-Z read anything, and had built like a vocal library of Jay-Z, and had him reading Shakespeare and other things. YouTube took them down for a couple of days, and then they appeared again, and basically YouTube was like, there is no copyright infringement here. We can't do anything about this. And so I don't know that they have figured this out yet.
Lauren: Are you optimistic about any part of this?
Puja: [Laughs] I mean, I'm optimistic about the opportunity that it presents to musicians who want to use it for good.
Gideon: What is using it for good?
Puja: So musicians who want to use the models to help their own work, um, that is still created by them. So I mean, I like the idea of artists being able to have a sounding board for their own work.
Gideon: Right. So there's AI as the kind of artistic amanuensis.
Gideon: I just had to throw that word in there, sorry.
Gideon: In some ways, it's the same story across all of these fields that AI touches. There are people who are good at a thing, and then they could use AI to become even better at that thing, or at least more productive at that thing. And then there are people who don't know how to do the thing, and they can use AI to perform a cheap imitation of the thing.
Gideon: I think the fear is that all the people who are performing the cheap imitations using AI will somehow crowd out a lot of the work being done by the people who have actual skills.
Gideon: And then there's a question of, well, maybe, maybe that won't work out. Maybe most of the cheap invitations are actually not that good. Not good enough. And so people with skills will net overall benefit, but I think a lot of it ends up coming down to who controls the tools and who has the power to implement them.
Lauren: But wait, if we can step back for a moment and I can just be philosophical. Why do we need an AI website that generates 14 million tracks and puts them on Spotify? Why do we need to be constantly improving? This makes me think of our New Yorker colleague Jia Tolentino's essay “Always Be Optimizing.” Why is it that, I mean, computers don't have limits. They do, but they exist to keep getting better and better and better. And we as humans and as human writers and editors and artists, like we do actually have limits. So like at what point do we just say, like, we can't always buy into the hype of this thing, that this tech tool is going to make us so much better and faster and optimize us?
Gideon: Well I think that’s driven by the logic of capitalism, isn't it?
Lauren: It is! Because the platforms and the labels and all of these much larger entities are ultimately what are going to make more money from this.
Gideon: Right? They're driving the optimization and the need to produce more and more and more.
Puja: I understand where your like existential, um—
Lauren: Angst? Crisis? Mid-podcast crisis.
Puja: Yes, because I share it. Have you heard of the Human Artistry Campaign? Heard about this?
Puja: So the Human Artistry Campaign is basically a collective of organizations across creative communities. It's like 60-plus of these, which includes actors, publishers, songwriters. It's a massive effort specifically tied to AI, because we're all in this existential spiral, right? And so they are kind of campaigning and lobbying and going to court to speak about the parameters within which AI should be considered as it pertains to the law. And all of that is tied to the basic return to the human coming first. And they have this set of principles that they are all kind of petitioning towards and different organizations have all signed on and agreed with, um, that are basically just like, “The law should be human first. Creation should be human first. The law should protect and require the consent and enthusiastic agreement of humans, and anything to do with any of this, including the algorithms that are made around these generative tools, should be transparent.” I mean, I think this is why, not to totally derail, but what is happening with the WGA strike and the one clause around AI is going to be very, very formative to the way the rest of all the creative industries are operating.
Gideon: Absolutely. I think it's a pivotal case.
Lauren: Yeah. Puja, I know that, um, we're asking you to make predictions on this podcast, and you indicated that you didn't want to. Um, but if, if I were to ask you—
Gideon: But you don't get away with it that easily!
Lauren: We still have a little bit of time here, so I'm going to ask you: Five years from now, you open up Spotify or Apple Music. What does it look like, not from a UI perspective, but what kind of music are we listening to? What's that experience like?
Puja: I mean, I work at Pitchfork.com. Let's use this as a caveat. I am constantly going to be enthusiastically advocating for the return to taste and the return to, to self-curation and being finicky about what you listen to, because it is literally core to my entire life. So I actually think there's going to be a rejection of the broad-brushing algorithm. I think people are getting more and more particular about having very clear, specific tastes, and also about not letting the computer trick them.
Gideon: So human taste is gonna be the vinyl of tomorrow.
Puja: I'm, well, the vi—you've seen this, right? Vinyl is booming.
Gideon: Right, exactly.
Puja: I really think there is going to be a return to something that feels a little bit more permanent, and tangible, and that has personal affection to it.
Lauren: My other podcast cohost, Michael Calore on the Gadget Lab. He is definitely, he's a musician and an audiophile in addition to being a WIRED editor, and we like to tease him on the show because his recommendation every week is like some obscure Swedish prog rock band we've never heard of. But I mean, I kind of love it too, because I'm so normcore, and I do listen to whatever Spotify serves up to me and probably a lot of pop music that I'm familiar with, and it's nice to have that, that tasteful human influence of “Try something you haven't listened to before, and I can vouch for it. It's good.”
Gideon: It seems, if you're suggesting that we're gonna value the thing that is created by humans, even if it's practically—
Gideon: Even if it's practically impossible to tell apart from the thing that's created by machines.
Puja: A hundred percent. And I will just say like, look at the way that ticket prices have moved in the last five years.
Gideon: Tickets for live events.
Puja: Tickets for live events, tickets for tours, tickets for concerts. There is such a desperation to be physically in a place, surrounded by people who feel intimately connected to the same thing as you. And having real personal access to an artist that people are paying insane amounts of money for. But I think it's also just reflective of this desperate want for human connectivity, and being boots on the ground and feeling something real.
Gideon: Puja, this has been an education. Thank you for joining us on Have a Nice Future.
Puja: Thank you for having me.
Lauren: So, Gideon, you know what I'm still curious about? What music do you listen to?
Gideon: Oh Lord. I find it really hard to explain.
Lauren: Open the Spotify app. Do it live. No cheating.
Gideon: Spotify also finds it really hard to explain what I listen to. Basically, I grew up listening to exclusively classical music, and I still go back to a lot of the early-20th-century composers, like Shostakovich. Um, I have been getting into more contemporary, minimalist classical, like Arvo Pärt and Nils Frahm and people like that. And then basically, if you match a country with a genre, like Ukrainian reggae or Nigerian funk or Vietnamese jazz, I usually like that.
Lauren: That's exactly what I would picture for you. Classical, but also beautifully esoteric.
Gideon: Aw, you say the sweetest things.
Lauren: Well, you are my boss.
Gideon: What about you? What do you listen to?
Lauren: I'm opening my Spotify right now, and I can tell you that. Um, OK. I was listening to Miss Anthropecene this weekend, as well as Art Angels, because I was trying to get into the Grimes vibe. I listened to the new track that was dropped by Beyoncé and Kendrick Lamar called “America Has a Problem.” And then this is, this is a little bit embarrassing. There's something called My Peloton playlist, because if you are on Peloton and you enjoy a song while you're working out and all those endorphins are flowing, you can just “heart” it on your Peloton, and then it automatically sends it to your Spotify and creates a playlist for you. So there are all these songs that I hear sometimes on Spotify where I'm like, why did I like this song? And then I realize it's because I was just drunk on endorphins while I was working out and thought it was a good song, and actually it wasn't.
Lauren: And that's, that's the Peloton playlist.
Gideon: So you are, you are biologically conditioned to your music. That's something.
Lauren: In some ways. Yeah. Well now I feel like we both have a little bit more insight into each other's psyches, because as Puja said, music really is a very personal and human experience.
Gideon: Yes, though I'm still trying to figure out what it reveals about me. I like the “classical but esoteric.” I'm gonna use that line. Anyway. If you want to have better music taste than Lauren and me, you should definitely give The Pitchfork Review a listen. That is Puja’s podcast, and new episodes come out every Thursday.
Lauren: Thanks for listening to our podcast in the interim. Have a Nice Future is hosted by me, Lauren Goode.
Gideon: And me, Gideon Lichfield.
Lauren: If you like the show, we'd love to hear from you. You can leave us a rating and a review wherever you get your podcast. And don't forget to subscribe to get new episodes each week.
Gideon: You can also email us at nicefuture@WIRED.com. Tell us what you're worried about, what excites you, any questions you have about the future, and we'll try to answer them with our guests.
Lauren: Have A Nice Future is a production of Condé Nast Entertainment. Danielle Hewitt and Lena Richards from Prologue Projects produce the show.
Gideon: See you back here next Wednesday, and until then, have a nice future.