Offline with Jon Favreau - Why The Music Industry Is Embracing AI

Episode Date: June 30, 2024

Chat GPT isn’t going to top the Billboard Hot 100 any time soon, but something is happening with AI and music—something’s BEEN happening. Unlike in entertainment and journalism, big music labels and even musicians like Drake and Grimes are cautiously embracing the latest in AI. And the results are not all bad! New Yorker writer John Seabrook sits down with Max to explain why the music industry has historically adopted new technologies, and how that Muddies the Waters around what is made by humans vs. what is made by machines. What does the future of songwriting look like with an AI Bob Dylan? Will a tide of lowbrow AI slop hurt artist payouts? And what’s really behind the record industry standing with artists? For a closed-captioned version of this episode, click here. For a transcript of this episode, please email [e-mail] and include the name of the podcast.

A lot of this stuff, people are just not going to want. Do they want to hear John Coltrane collaborating with Carly Rae Jepsen? Is that going to seem ghoulish and weird? I don't think we know. I'm Max Fisher, sitting in for John this week with something I've been wanting to bring you for a while. I am pretty skeptical of AI's potential in creative fields. The major AI models are nearing a technical ceiling while their output remains stuck in the uncanny valley.

This technology was supposed to bring us AI-generated novels and feature films, and instead it's drowning us in deep fakes, weird spam images, and sloppy forgeries. Music might be an exception, though. I'm not saying that ChatGPT is going to top the Billboard 200 anytime soon, but something is happening. In other fields like entertainment and journalism, people are rebelling against technology that they perceive as a threat to not only their jobs, but to everything they value about their craft.

And that is not quite what you are seeing in music. Yes, there is skepticism, there is concern, but the big labels are cautiously embracing AI. So are a number of artists, and the results are not all bad. Here to talk to me about why this is happening and what it means for the future of music is journalist and author John Seabrook. John has long covered the music business and many other topics for The New Yorker. Most recently, he reported deeply on that industry's experimentation with AI. John Seabrook, welcome to Offline.

Thanks. Great to be here. So before we get to AI, I want to talk about Napster, the file sharing service that launched in 1999 and almost killed the music business. Because I think this is really important for understanding why 25 years later, the industry is responding as it is to AI. So can you talk about how people in the industry today think about that Napster episode when they look back on it and what lessons they took from that? Yeah, a lot of people sort of talk about what's happening today as like a Napster moment for the industry. But it really isn't a Napster moment because there's so many differences. So let's just go back to the actual Napster moment, which was the RIAA sued people who were using Napster, you know, users, ordinary people. And that lawsuit is the thing that people are kind of comparing the RIAA lawsuit to today. Because, yes, RIAA is suing Suno and Udio. But in the former case, it was a very unpopular lawsuit that the labels were behind it,

but they were sort of very quietly behind it. You know, artists that actually did support it, like Metallica, got crucified by their fans. So, you know, and for the fan, Napster actually did represent something of value, even if it was illegal. And that was that they were getting music that they had to pay $20 for a CD for, you know, for one song that they wanted to listen to. They were able to get that song for free.

So there was a real, you know, economic advantage to it, even were able to get that song for free. So there was a real economic advantage to it, even if it was illegal. If you look at AI and today and what Suno and Udio are offering the user, it's not like free music or you can get free music now from, if you don't mind the ads. So it's not as clear a sort of value proposition. And the other big difference is almost everyone is united behind this lawsuit. The labels, all three labels are behind it. A lot of the artist representatives and the artist rights groups are behind it. Basically, instead of this kind of oppositional relationship. And back in the Amsterdam days, the labels had very different ideas about how to deal with it. And a lot of them sort of set out on their own to try to do it without working with the other label groups.

And that's very different today, too. The label groups are working together and they're really kind of working for the artists because, you know, the artists can't sue Suno and Udio. They don't have the wherewithal to do that. So they need, you know, these organizations to help them. So, yeah, there's a little bit of context with Napster, but I think the differences are much more sort of profound than the similarities. Oh, I completely agree. Very different kind of technological disruptions. And I take your point about the legal strategies and the similarities there. I think what was really striking to me about your piece when you were interviewing music industry executives today is how they look back on the Enabster episode and how it is informing their thinking about AI today. Like you interview Lucian Grange, who's the head of Universal.

And he has this quote where he's looking back at the Napster era and thinking about how that informs how he wants to deal with AI. And he says, quote, if we had waited for the tech companies to finish innovating, we would have ended up as roadkill. And indeed, a lot of the music industry did end up as roadkill out of the Napster era, where there's this kind of fork between the companies that said, we're going to resist change, and we're going to embrace it. So how is that, do you think, today informing their broader view of how they think about AI? Right. I think a lot of people took the lesson from Napster that there's no point in trying to put your head in the sand or sue them out of existence and use only lawsuits as your strategy because it won't work.

And, you know, the tech is inevitable and it's coming. And the younger people who were around when illegal file sharing started to cut into the industry, people like Lucian Grange, who was, you know, kind of earlier in his career, saw that very clearly, whereas older people who were in this industry then sort of didn't embrace change, tried to resist change. You know, technology had kind of been the friend of the industry. You know, CDs had been a huge economic driver for the industry. And so I think people sort of thought that technology was kind of their friend. And then Napster came along and it seemed like it was a different kind of tech world

altogether. So yeah, I think what you're seeing with Lucien and some of the other labels, as quite opposed to the film studios that have fought, you know, people who like fought actors over AI. You see the labels actually working with the artists and trying to, I mean, you know, parts of AI are great. And so it's not like all AI is bad. And they're trying to sort of pick out the stuff that is good and can work with artists and their business model and not undermine it.

I want to come back to this idea that the music business is approaching AI very differently than other industries like films. I think it's a really important point for understanding the role of AI in music today. But something that you kind of touched on about the music industry adapting to the rise of file sharing, like, you know, I always think if you had asked me 20 years ago to pick a creative industry that would be most imperiled by the internet, no question i would have said music the file sharing cut their revenues by 40 which is crazy piracy was rampant and easy you

could find any song free to stream on youtube and yet today and this is something you emphasize in your piece the music industry is the standout success of the internet era in terms of their revenues in terms of how locked in they are to the business, in the sense that it's like really profitable. So how did that happen? How did they make this transition to actually ride the wave of the internet instead of being overrun by it? They had a really good model in Spotify. They had a model that had been proven in a smaller market, which was Sweden, which had been absolutely decimated by privacy. I mean, sorry, by piracy, privacy. You know, the industry executives in Sweden were basically just sort of wringing their hands and giving it up and really just tried Spotify because they had

absolutely no other choice. And then, so then the U.S. industry was able to sort of see how that worked and, you know, a couple of years and then, you know, in a slow sort of incremental are beginning to fall i think you're seeing more criticism of the free side of these services but but the freemium model really worked as it was supposed to work which was to take people who were used to getting their music for free and offering them you know uh better services and better labeling and you know all the other other things that Spotify and the others offer, and converting them to paying subscribers. So it had a really good working model, and it converted people. There isn't really something like that yet for AI, although maybe we're going to get there. I mean, Spotify was licensed in 2000, I think in 2008 in Sweden, and maybe 2010 in the U.S.

And, you know, Napster appeared in 1999. So it took, you know, 10 years to get around to Spotify. But, you know, it turned out to actually be a working model. Maybe not perfect, but it allowed the labels to transition from a point-of-sale product model to an all-access streaming model in a way that added to the revenues pretty quickly. think that I was really struck by that I don't think I had fully appreciated the kind of audacity of the music businesses shift to streaming and how early it was relative to other industries where I think there was someone you interviewed that actually talked about the music industry had forever and ever been a wholesale retail business you sell a physical product to people. And they basically abandoned that. They basically went to this model of you buy a subscription for access to our catalog, or you listen to ads and that gives you access to our catalog. And it kind of feels like no

other creative industry has made that big of a shift into the internet era, right? No, they haven't. But I mean, it's partly because I think the nature of music, the small sort of bandwidth size allows you to stream from so many devices, so many different kinds of platforms. And then social media kind of came along at a similar time and really helped drive streaming online, YouTube and other online services. But if it had been video or film, I think it would have been technically much harder to make it work

because the packet sizes are just so much larger. But the music industry is kind of the leader here and the film industry. I mean, the book industry is actually another interesting example. But it may be that those industries follow. But I think it was partly just the technical aspects of music, the way people can use music in many different settings. It can be streamed while you're doing a lot of other things. And it just kept growing in terms of the revenue streams that streaming generated, more and more platforms developed and they were able to license them,

the most recent being TikTok. And so you got one revenue stream adding onto another, add onto another, and the economy of the music industry has just continued to grow. Yeah, and that's a good point about the technical nature of music does make it more adaptable to this streaming

internet era. And at the same time, this is something I really hadn't understood until I was reading your work is that a lot of the music executives who lived through that shift and who lived through the Napster era, they talk about it like they've lived through World War I, which it kind of was for the industry. And they have this kind of like, never again, kind of mentality. And TikTok is actually a this kind of like never again kind of mentality and tiktok is actually a really good case study for this could you talk a little bit about how they adapted to that because again this is something that could have been a huge threat to the industry but maybe partly because of the lessons they learned from the napster in the file sharing era

the music industry actually makes a ton of money off of TikTok, right? It is now. But, you know, only until relatively recently. You know, TikTok was basically not paying anything or very, very little in terms of royalties, much less than YouTube, which is still a point of contention, doesn't pay a huge amount. But no, the feeling was that TikTok wasn't paying their share at all. And so recently, Universal, they withdrew their artists from TikTok and said, okay, see if you can make it without all of Universal Music Group's artists, which represents more than a third of all of the recorded music out there. And no one really liked that. TikTok certainly didn't like it, but a lot of the artists didn't like it because they used TikTok to promote their music. And so when Taylor Swift

basically kind of broke the Universal ban on TikTok and started promoting her most recent album on TikTok, then Universal realized that the game was up and they did come up with a licensing agreement with TikTok, which was a lot better than what they had before. And as I've told, better than the licensing agreement they have with Google and YouTube. It's also important to mention that Google is an interesting player in all this because Google has been like a good guy in this. A lot of my piece was about the product that YouTube and Universal came up with to allow YouTube users to use voice clones of Universal artists. And now there's a new product Universal has for that. has been sort of an exception to the way the tech industry has typically sort of dealt with the music industry,

which is like, yeah, as you were saying earlier, this is an industry that needs to be disrupted, because it does seem like kind of this kind of cottage artisanal industry. You've still got these guys that are making choices on who's going to be a star that doesn't really have a lot of data or science behind it. And so if you're a tech person, you just say, come on. This is the 21st century. And yet it's not as easy as all that, apparently, to make music. Anyway, so that's another piece in this whole thing.

I think Google has been a kind of intermediary between the content industry and some of the other tech companies. Well, Google is a really interesting case study. Google, of course, owns YouTube because initially YouTube was like the next major threat to the music industry, right? Because you would see all of these music videos and songs that pop up on YouTube. It was free to stream. And like you described the music labels doing more recently with TikTok, they kind of took this two-prong approach of like, we are really going to fight you legally and in court if you are going to try to stream our music without paying us. And we're going to give you takedown notices. We're going to get really aggressive about litigation. But at the same time, we are going to try to bring you and work with you to strike these licensing deals so you can have the music but you're going to pay us for it. And YouTube pays – isn't it like $6 billion a year or something to the music? It's a huge amount. Yeah.

It's a lot. Okay. So they followed a similar strategy with TikTok, which seems to work really well and seems to fit in this kind of strategic approach of like, we're going to embrace the internet. We're going to embrace this disruption. We're going to own it and bring it into the business model. But we're also going to like really use our collective economic and legal power to coerce these tech companies to make sure they play nicely. So can you talk about how that is informing their approach now to AI? Because it reads to me as kind of similar to that. Yeah, I mean, you do have to remember

that the content companies like Universal, even though they're huge companies, are minuscule compared to the market capitalization of the tech companies. So there's a huge imbalance in terms of just the sort of economic might of the two sides. And it takes a lot of fancy footwork on the side of the universals of the world to get these deals. And so in the case of Suno and Udio, the three major label groups are all suing together through the RIAA, the Recording Industry of America's kind of lobby group, that is suing not for copyright infringement per se, not on the outputs of what Suno and Udio are making, although you could sue on that basis. But on the input, they're saying that

copyrighted materials were used in training these models, and they're asking $150,000 for each case, which would, you know, if you added it up together, would be like an astronomically large number. So they're going after both the companies themselves and some of the investors in the companies, or at least they're opening to the possibility of going after investors. And again, as with TikTok, I think they would be very happy with a licensing deal with Suno and Udio. How you would do it is a big question. Whether you just sort of license the whole training database and any artist who's part of that training database gets sort of the same amount.

Or whether you look at it like it sample like as like as a sample the way sampling works uh and sampling is actually also an interesting uh parallel here because sampling started out with people just taking stuff and putting it into records and then there were lawsuits and then there was a model worked out for equitable payment, and now you pay for samples, but you pay on the output side. Like if you use however much of the sample you use, you're going to pay more or less, depending. And so if you try to take that and make it part of this UdioSuno output, you need an AI to figure out what part of Prince is in there and what part of the Beatles is in there. It's kind of mind-boggling to think about how you would figure out the rights to that. But anyway, that's the basic approach. It's either going to be

licensing or there's going to be a settlement. Sun know, Suno and Udio would settle, or they would go to trial. And so if they go to trial, that would be Suno and Udio claiming fair use, which we could talk about as well, but that's basically their defense, that they have a right to these songs to train their models with under the doctrine of fair use. I want to come back to this lawsuit because I think it says something interesting about how the industry is approaching the kind of low end of AI. But I want to talk a little bit about the higher end, the stuff that sounds a little bit more polished. And a big moment for this and for the industry seems to be the deepfake song that got published in April of last year with an AI that was mimicking Drake and The Weeknd. And honestly, it sounds kind of passable.

I think we have a clip of a few seconds of that that we can play. Okay, it's not bad. So can you talk about what the reaction was to this in the industry and including among investors, which seems to have been a big part of this? uh, renumerative aspects because, uh, you know, let's just think about what an artist who, who has a, an AI of his voice could achieve with that. Uh, you know, you could basically let people write songs with your voice and then you could

decide whether you like them or not. And then you could be a co-writer of that song. Or if you have a hit in English and you want to have a hit in Swahili, you don't have to find a Swahili singer to sing the song. Your own voice can be used to sing in Swahili. And that's another, you know, potentially large economic driver for this. So, and also there's a different set of rights involving the voice. It's not actually copyright. It's what's called right of publicity. They're less defined as copyright, and they're not federal laws.

They're state laws. So there's a whole sort of different area of this sort of larger movement everybody up, literally like overnight, it seemed like, to what could happen if the industry doesn't get involved. And it immediately sort of drove a number of projects to get involved. Again, because voice cloning is, in the music industry, an artist's voice is such an intimate part of their artistic soul that I think it hit people on a visceral level. And that's also part of the Suno Urio lawsuit,

is that the voices that these models generate, you know, they do sound very close to some of the artist's voices that are used in the prompt, even though it's not supposed to work like that. So the voice part of it is, you know, a real driver of a lot of the sort of panic, I think, or the anxiety. But also, you know artists, like, I think, I mean, even Drake and The Weeknd's reaction to that wasn't necessarily negative. I mean, Drake said something funny. I can't remember. He said something like, they've gone too far now or something like that.

But, you know, I think, and this Drake BBL thing that was generated with Udio is a similar thing that became a huge hit. And so there's clearly the potential there. And the industry wants to get a piece of it. And artists are worried about being protected. But, you know, the artists do have the rights to their voice. That's clear. No one's disputing that. So I think they're in a very good position.

Yeah, and it's a good point about, or it feels like an important point about it hitting artists very viscerally because it also hit investors very viscerally, right? There's a point you made that Universal stock around the same time dropped by 20 percent and this is for a number of reasons but partly was concerned over generative is this a new threat to the music industry and part of the industry's response to this big wake-up moment was to as

you said to try to like proactively lean into initiatives or like how can we own ai and bring it into the business model the same way that we did with streaming, the same way we did with the internet? And there's this scene in your story about one of those initiatives formed kind of in response to this, that I find myself recounting all the time to people because while I do remain broadly skeptical of AI's creative uses, this story added a layer of kind of nuance to my thinking on this for me. And it's about Don Was, who is a celebrated music producer,

now the president of Blue Note Records, playing with this Google AI music tool called Lyria. So can you describe what happened and what his reaction was to that? Yeah, so all of the big AI, the big tech companies have their own models, right? Lyria is, you know, Google's, you know, music model, which is, you know, essentially, it's like their AI that's sort of trained on music rather than trained on images. It's, you know, the core tech with a music training on top of it. And so, yeah, so Google started this artist incubator program to allow music professionals to play around with Lyria, which is a generative AI product that you can create songs and lyrics and artist voices. And so it did that.

And Don was one of the people in that incubator. And Don, who is, as you said, a great musician, great producer, produced a lot of Dylan stuff 20 years ago. And he is the head of Blue Note, the label, the classical, the jazz label at Universal. So has a foot in the creative world and a foot in the business world. He played around with it. And he told me, you know, it was like, he said, his phrase was, it was like collaborating with himself on his best day.

That essentially it was like a co-writer, you know, a lot of music is written in groups of two or more. Something about co-writing seems to be particularly useful when you're creating music. And so, but you don't necessarily have your co-writer. So this was a co-writer that could be trained either in your own, you know, on your own work. So it's sort of you, you're collaborating with, or Don also speculated it could be trained on a famous artist's work, like, let's say, Bob Dylan, so that, like, you could pay to co-write with Bob Dylan. And Bob would get paid, you know, for that. And I don't know if that's really going to fly because there's going to be a lot of a collaborative um thing that gets better as you work with it

that has access to your entire oov and you know whatever other oov you feel like dipping into uh so the ultimate sort of synthesizer workstation digital works you know a lot of people see tech, the AI as a continuum. Synthesizers, drum machines in the 80s were also hated by a lot of people in the music production business. They were seen as, in some ways, some of the same language you're hearing about in AI music today, sort of soulless, inhuman sounds. But, of course, they were embraced by people like Prince and David Bowie, and then they became mainstream, and one expects something similar to happen with AI music-making tools.

And so, yeah, that's what Don was excited about. And not just Don, but the other people in the incubator were excited too. How that's actually going to play out as a, you know, whether musicians are really going to make music with AI that's, for them, a more creative experience than before. I haven't talked to a musician that's actually had that experience, but in terms of a song that came out of that, an AI collaboration, but it certainly sounds like a good thing. Unless it learns to do what you do better, then you can do it,

and then it doesn't need you anymore. That would be the negative outcome. Well, this is sort of what I found to be really interesting about the reaction from a producer like Don was, is that he's not saying this is replacing me because it can spin up a Don was out of thin air, but rather that it can train on his existing material and things that he has actually done that he innovated that came out of his brain and his creativity. And that it like you said that it felt like collaborating with himself on his best day.

And he has this quote, the songwriters are going to like this more than anybody as long as you can't steal from them. And I want to ask if you get a sense for, is that a common view among artists, do you think? Is there a sense that this could be an exciting new tool kind of on my workstation? Or do you get the sense that people are more like, this could be used to replace me? You know, a lot of that is age-specific, I think. Like, you know, Grimes, you know, a younger artist,

obviously has really embraced AI, has basically put her voice and her her song stamps out there and invited collaboration with people and and um it's a different relationship with your fans i think it's uh and you see a little bit of this on tiktok too how sort of allowing fans to remix your songs to slow them down or speed them up or, you know, kind of mess with them and being cool with that and inviting that. That's maybe a younger or it's a certain kind of creative person's attitude. But there's our other creative person's attitude that doesn't want to collaborate with the fans, wants to really make their work perfect and finish it and then put it out there.

And, you know, I think like a Florence Welsh or someone like that. And those kind of artists maybe aren't going to be as quick to embrace AI. I think, I mean, the difference between, you know, musical artists and actors in terms of AI and their relationship with their studios or their labels is that studios are basically saying to actors, you're just sort of work for hire. We can use your voice any way we want. And you don't have the right to say, you know, you have control over your voice. That's very different from what's going on in the music industry, where basically the labels are saying, we do want you to be able to make money from these AI products, but we are going to protect you, and we're on your side. And that's, you know, it's really different from the film industry,

and it's actually kind of unusual in the context of the music industry where labels have traditionally been the bad guys, right? Like, they're the man, and they have a long record of fucking people, you know, going back to black musicians who were completely ripped off. So it's not like, you know, they haven't engaged in terrible practices, but they're in this rather unusual kind of maybe do found role of being kind of like the good guys here. So we'll see how that plays out, but it's different. I think that's a really important point that so much of how this is going to work is going to end up turning on the kind of public sense and the industry sense of the relationship between artists and labels here. And it's a really good contrast, I think, with film and frankly,

with journalism too, where there's just a norm has been established. And I think with some good reason that AI is something that the movie studios want to use or that, you know, in journalistic industries that the companies want to use to cut costs and that this is something that is bad for the creative professionals. And, you know, the fact that we saw the Screen Actors Guild strike around this time, WGA strike around this time, fighting specifically against the use of AI has really drawn these battle lines where, like, AI is bad for creatives is the narrative. And it's good for the bosses. And that's where the kind of push and pull is going to be. And it does seem like at least so far, it is kind of an amazing reversal, as you said, that the music labels have positioned themselves as actually protecting the artists in AI by saying, we're going to bring it into the business model.

But I wanted to ask you a little bit more about what your sense is, to the extent that it's possible to kind of get the temperature of people on this of the artists on maybe their view of other artists who are dabbling in this like one venture that you write about is this thing called dream track and correct me if i'm misstating this but my understanding is that as a collaboration with youtube or youtube creators can use ais that have been trained on certain artists where it's like i I think, John Legend, Charlie XCX are two of them. But it's not really those artists like John Legend creating a track using AI. It's YouTube creators that are using AI trained on those artists to make videos. And do you get a sense, like, do other artists see this as just like, kind of a gross cash grab? Or do they see it as something that is interesting and exciting?

I think most artists see the product that you were talking about, which is called DreamTrack, which you described accurately. You know, I think that's like Google trying to figure out how they're going to make money on this. They're trying to figure out how to turn it into a product because, you know, they've invested all this money in AI and they need to figure out products. That, I think, that part of it sort of seems like a sideshow probably.

I mean, I know even to the guys at Universal on the label side, I mean, I'm not going to characterize their attitude, but I think a lot of them go into something like that sort of holding their nose, thinking like, okay, we'll do this, but then we'll get the incubator out of it. And so it's kind of a horse trading kind of situation. But I do think a lot of artists are still very leery of big tech, for sure. I mean, particularly the ones that live through the streaming area and streaming

era and saw their CD sales collapse with, with streaming and never really come back. And, and, you know, guys like Daniel Ek just saying, we're going to do this, you know, without really asking. And, and then they sell their do this, you know, without really asking. And then they sell their stock for, you know, millions. So, yeah, I don't think a lot of artists are all like whistling, you know, going along with the tech companies on AI. I feel like there's a certain sort of desperation even like, oh my God, not again. And really realizing that they have no real allies here. I mean, the labels are as close to an ally as they have.

And then there's, you know, like the human artistry campaign. And the other part of this, I guess, is like what does the public think about songs that are made with AI? And will there be like, you know, the way sort of like a labeling, like GMO corn type backlash, where it's really going to fall to the public to say, I don't want to listen to music that's made with AI just because I don't feel like it. And I want my music, you know, sort of, you know, made by humans. That's going to be a really hard distinction probably to draw because humans are going

to want to use AI to make certain aspects of their music. So how do we sort of define what's human made and what's AI made is going to be complicated. But I do think labeling is a possible way forward, and I think the public is open to wanting to know whether their music is made with an AI or not. And ethically trained AI versus unethically trained AI is going to be a growing distinction too. Well, there's another episode you recount that I think is so fascinating that speaks to this exact point you're making

about what, how is AI going to be pulled into music? What does that actually look like? How is that going to work? Where you talk about Neil Mohan, who's of course the CEO of YouTube, playing a demo of, I believe it's also this Lyria tool, an AI demo of Carly Rae Jepsen's Call Me Maybe in the style of John Coltrane. He's playing this for Universal top executives. And some of the executives are nonplussed because they're saying artists are going to fucking hate this. But Lucian Grange is the CEO we mentioned, compared it to, and you kind of hinted at this, the rise of synthesizers and drum machines in the 1980s, which were things that, like you said, were initially viewed as potentially a threat to musicians

because, you know, is it going to replace a drummer with a gadget? But now they are viewed as, you know, they're celebrated as a really important part of music because otherwise, without those tools, you wouldn't have Daft Punk, you know, you wouldn't have Superstitious. Artists ended up using those tools not to, like, cut labor costs by firing the drummer, but by leaning into the ways that those tools sounded unnatural and by expanding what they could do creatively with that. And that made me wonder if there is something maybe different about music than other creative pursuits, other creative industries that makes it more adaptable to technological disruptions to the creative process? Well, I mean, one thing I just keep coming back to is the fact that the revenues, like music from 50 years ago can be repackaged and resold and listened to

by completely new audiences that may not even know that it's old music. The role that catalog plays in the whole economy of music, if you compare that to the role of catalog in film or the role of catalog in books, it's just a much more essential part of the ongoing economy and the mix between old and new. You mentioned the sort of Carly Rae Jepsen in the style of Coltrane. I mean, this is another thing that AI promises. Maybe people don't want it, but you can have collaborations between living artists and dead artists, between dead artists.

You can have work of dead artists that was never recorded that can be recorded and sung in their voice. And this also goes beyond music. I see that NBC is going to use Al Michaels' cloned AI voice to commentate on Olympics events at the Olympics that he's not even there to comment on. You know, I mean, so it's kind of happening. But a lot of this stuff, people are just not going to want. Do they want to hear John Coltrane collaborating with Carly Rae Jepsen?

Is that going to seem ghoulish and weird? I don't think we know. It's going to take a lot of stuff, sort of throwing stuff out there. But at any rate, to go back to your original point, I do think that the music industry's, the fact that old music, and also old music sort of sounds a lot better than music because it was recorded with a different process and a much more like with real professional studios and musicians and and now you get dolby atmos sound and new ways to listen to it and ai's allow allowing you to do things like that Beatles,

you know, the new Beatles song where they actually took the stems and remixed them and used AI and the whole Peter Jackson Beatles documentary. I mean, those are situations where AI is being used to like clean up old recordings or bring out. Like I listened to, when I was doing my Universal piece, I listened to an Eleanor Rigby remix using, you know, the latest AI. And I heard stuff in Eleanor Rigby I'd never heard before. It was fantastic.

So that exists for music in a way that I'm not sure if it exists in other industries. I mean, I also wonder if there's, It's really fascinating about Eleanor Rigby. I wonder if there's also something about the, like... There's a synthesizer in there that you don't really hear. In Eleanor Rigby? I'm telling you. Yes, yes, there's a synthesizer in there. Wow, I had no idea.

Anyway, let's talk about the, like, lower end of the spectrum of AI. In any field, there is a lot of AI slop, a lot of AI garbage out there. Spotify and YouTube are getting flooded with more and more AI music. A lot of it is basically white noise for listening while you sleep or you exercise. But some of it is pushing into the cheap knockoff territory. And these are the lawsuits that you mentioned earlier against these two AI startups. And we have a clip that we can play of Mariah Carey AI knockoff

from one of these. And this clip, I should say, was actually generated by the lawyers for the music labels who attach this as evidence of their lawsuit against the companies whose models produced the song. They say we're illegally trying to copyright music. Let's play a little bit of this. Okay, wow, I hated that.

So there's part of me, if wonders, that this stuff, like this scalable AI slop, is maybe a bigger threat to the industry than the passable Drake deepfakes. But how does the industry view this? Do they feel like they have this under control or are they worried about it? No, they're worried about it. I mean, part of what that does is it dilutes the royalty pool i mean the way because of the way that the streaming economy works artists are paid on a sort of proportional basis what the number of their streams relative to all the streams in a given month so when when you increase the streams with AI and you have hundreds of thousands, millions of tracks flooding onto streaming services, that's decreasing the amount of

money that an artist is going to get for their particular streams. And that was actually a big point of contention in the TikTok universal dispute. TikTok wanted the right to be able to put AI music up on TikTok and count that in the royalty stream. It's basically royalty-free music that really only serves TikTok users, but it also hurts the general royalty pool for legitimate artists. So that's one way in which AI could, like a tide of slop could directly hurt artists' payouts. But then for the listener, of course, I mean, discovery becomes all the more difficult to do,

which is already sort of a big issue. For professionals who do make sound, a lot of people work in the industry and don't make money as pop hit makers. They might be making commercial music, and often in commercial music, you're asked to compose a sound like Mariah Carey's Christmas song, but make it different enough so that we don't get sued.

And there's a lot of people that do that stuff, and now all those people will be out of work. So that's another aspect of it, too. And the final point I'd make is Mariah Carey's Christmas song is actually getting sued by a human now for copyright infringement. So it's not just like, you know, AIs that are doing this stuff. There's still good old, you know,

infringement going on among humans, and that's probably going to keep going on too. It's a beautiful tradition that will never die. I mean, I guess the good news is that the labels at least have a pretty good track record of being able to bring these new technological challenges like streamers kind of like to heel to some degree and bringing them inside the tent. And they do seem to have a view that the ultimate asset they want to protect is their catalog, which is to say that they see like human generated,

good, high quality music that can be licensed off of in whatever ways to AI is the ultimate thing they want to protect. So they don't want that washed away. And it is so ironic that this is all happening in an era where artists had been getting more and more independent from labels because they don't need labels for distribution or marketing in an era when they can promote themselves on TikTok, upload directly to Spotify. Now, suddenly, they do need the labels. And you mentioned this. They need their kind of economic and legal muscle to wrestle all of this out with the AI companies. So do you see any indication that a desire for protection from AI or protection from disruption is driving independent artists

back towards major labels in any degree? Or does the appeal for self-publishing and self-promotion still kind of prevail for people? That's a great question. I would say that really, if you look at the sort of trend of the versus today. It was like 80% then and 20% now. It's really changed in a huge way because of access to streaming and all the DSPs allow you to basically put your songs up. I mean, people don't listen to them. I mean, most of the songs that get put up don't even get one spin and the vast majority of them only get a few, but

that's a giant driver. So I think that the lure of independence and the ability to do it yourself is the strongest driver right now in that kind of label versus independent sort of world. But at the same time, you're absolutely right that being independent doesn't give you any protection against AI. There's nobody to really fight your battles for you. And also, it sort of depends on what kind of artist you want to be. But I think what the labels have done is basically said, if you want to be, what we can do is make global superstars. We can create Olivia Rodrigo or Billie Eilish or, you know, Taylor Swift. And if that's who you want to be, then you really do have to sign up with us because no one else is going to be able to provide that kind of support.

And that's probably kind of true. But aside from that, and it does seem like there are a lot of artists that don't necessarily want to be that kind of artist. You know, there's now a lot of talk in the industry about you know who's going to be the super bowl uh halftime star in 2040 because it's sort of looking right now that you're not seeing that kind of mega star sort of emerging you know in the next tay Swift. I mean, maybe Sabrina Carpenter, but too early to tell. At any rate, it does seem like the landscape has shifted away and there's all these people that are content making and are able to make a living. And the other piece of this is

now there's this creator economy that you can monetize your career with advertising or with influencer deals from the beginning and not need major label support. So there's a whole sort of now separate economic path you can follow to realize your dreams as a music person. So, you know, yeah, it's a really interesting time because AI does require sort of major label help if you're going to stand up for your rights. But the allure of doing yourself an independent and the ability to to to monetize is growing and and i think more and more people are kind of taking that route it's a fascinating era that we're in and i find it really helpful to bring myself back to what is happening in music with ai just as a way to remind myself that well i don't think that's going to be the path for a lot of other creative industries there are ways that this can happen that is a little bit more or that in theory hopefully are

a little bit more positive some for artists for the industry and certainly for consumers and for music fans and as a music listener i am uh you know 10 trepidatious but 90 really interested to hear what's going to come so john thank you so much for taking the time. It was great to chat. Super interesting. Really great having a conversation about it. Thank you for having me. I appreciate it.

If you want to help John, John, and Tommy make the New York Times bestseller list, the window is closing and today is the last day sales count towards that list. And if you're not persuaded by the fact that expert advice and tips from the smartest people in America are in the book, well, the chapters are also short with illustrations by George W. Bush and some funny jokes that are not by George W. Bush. Sounds perfect for those of us who struggle to make it all the way through a New Yorker article. Oh, I love New Yorker articles. Well, I kind of like this book too. Please help them get on the New York Times bestseller list. If not, John will be sad. And do you know who will have to pick upseller list. If not, John will be sad.

And do you know who will have to pick up the pieces? No, not John's wife, Emily. It's me. Are you comfortable putting that on my shoulders? Head to crooked.com slash books to grab your copy now or wherever books are sold. Do it for me or for democracy, whichever is more important to you.

Offline is a Crooked Media production. It's written and hosted by me, John Favreau, along with Max Fisher. It's produced by Austin Fisher. Emma Illich-Frank is our associate producer. Mixed and edited by Jordan Cantor. Audio support from Kyle Seglin and Charlotte Landis. Jordan Katz and Kenny Siegel take care of our music. Thanks to Ari Schwartz, Madeline Herringer, and Reid Cherlin for production support.

And to our digital team, Elijah Cohn and Dilan Villanueva, who film and share our episodes as videos every week.