1. Inside the Black Box

From Windhill Studios, this is Lightbox. The first time I played around with an AI Music Generator, it caused an existential crisis. I literally lost sleep, staring at my ceiling in the middle of the night, wondering, "Will this be my last year making money as a musician?" So, what happened? I was going a project that was Alfred Hitchcock-esque, Hitchcockian. I was supposed to write a dozen tracks that sounded like they could have been written by Bernard Herman, one of the greatest composers of the 20th century in a frequent collaborator of Alfred Hitchcock. "Great," I thought. "This'll be easy." Before I sat down at the piano for my impossible task, I thought I'd do a little experiment. So, I logged on to an AI Music Generator called UDO. There was a little box that said, "Describe your song." I typed in Bernard Herman, "Theme," Alfred Hitchcock, "Film," "Misterius," and then clicked "Create." About a minute later, I got this. My mind was blown. The composition, the orchestration, the recording quality, it was all spot on, a million times better than anything I could do. That's stunning. This is Michael Spitzer, author of The Musical Human, a renowned musicologist and Beethoven expert. I played him this track to see what he thought. So, that's AI? That's AI. It creates the actual sound, or. Yeah, all I typed in was words, and it produced that in about a minute. I'm still trying to process it, it's mind blowing, isn't it? And yeah. My name is Mark Henry Phillips. For the past dozen years, I made a living as a composer. Music for commercials, for clients like Google, Nike, and Ford, scoring indie films, making music for podcasts, podcasts like cereal, startup, this American life, and homecoming. It isn't exactly what I dreamed of when I was a kid, and I decided I wanted to be a musician. But on most days, I counted myself lucky. I was making a decent living by playing music. But now, it felt like the jig was up. This whole crisis started about a year and a half ago, when I logged on to Udo for the first time. I was there just to see what it could do, and I stumbled across a track by a user named Man or Monster. He was trying to recreate a Toots in the Middle's track, and I typed in some lyrics along with soulful reggae, ska, 1969, ham and organ, and out pop this. I didn't want to admit it, but it's good, really good. It doesn't feel like it was created inside a data center. It feels like it was made in a makeshift studio, in Jamaica, on a hot summer night in 1969. It brings up all sorts of copyright issues, because this sounds a lot like Toots, aka Frederick Toots Hibbert, aka a real live human being, and I would wager to bet that's because Toots' entire catalog was used in Udo's training data. But let's ignore that for now. We'll get to it in future episodes. For now, let's just focus on how good this sounds. It might not be perfect, but it feels like real music. And if it feels real, that has big implications. Like what you ask, let me give you a hypothetical. You're an ad exec, making a beer commercial. You want to track that sounds like it was recorded in Jamaica in 1969, but you don't want to deal with the money and legal back and forth that comes with licensing a vintage track. You might turn to a composer, like me. You pay me, not as much as you pay Toots or Desmond Decker, but still enough that it'd be worth my while. Boom, you get a commercial, and I have a job. Years ago, that's exactly what happened to me. I was hired to make a song that sounded like an early rock study tune from 1969 Jamaica. Essentially, I was given the same prompt as the AI track, but here's what I came up with on my own. It took me a couple days and it's pretty good, fine, mediocre. It's kind of embarrassing playing this. And I was going to take out the vocals because that part is super embarrassing, but it's kind of important to our story. [Music] Whatever you think of my track, the vocals on the AI version sound way better in that legally and morally dubious type of way. And don't get me wrong, there are many musicians out there who could have produced something way better than me and the AI. But the vocals? Here's the AI-generated track again. [Music] To get a vocal performance like this, you'd kind of need to be an amazing Jamaican legend with the group of amazing backup singers. And even if you were, you know, toots himself, you couldn't write, produce, record, and mix a track in 10 seconds. That's the crazy thing. Our hypothetical ad exec can now make this track. Help 10 tracks in under a minute. And it's basically free. See why I was having an existential crisis? This was my thing. You could give me a prompt, make a song for a beer commercial, whatever it was, and I could make you a song that fits. Not a lot of people can do that, so it gave me a job. And it also gave me an identity. And now a machine could do it in 10 seconds. And yet, when it comes to AI music, all I hear is that it's slop. It's taken as red that AI couldn't possibly create good music. So what was happening? Maybe I just stumbled across a few standout examples? I kept digging. And I kept coming back to soul music because, well, the title of the genre kind of explains it perfectly. Soul music is supposed to have soul. Surely AI couldn't do anything that resembled the holy grail of soul for me, Booker T in the MGs, the legendary house band of stacks records. But without really trying, I created this. I've listened to stacks regularly for decades. And if someone played me this track and told me it was an outtake on a Booker T box set, I wouldn't have flinched. You might be less convinced than me, but would you really call this slop? So why this huge disconnect between the slop narrative and my experience? Do I just have horrible taste? It's definitely possible. But I've made a career as a commercial musician with very modest musical ability. In some ways, all I have is my taste. So how could my taste be so out of whack with this general narrative? Something was going on. Unsure what to do with all these anxieties and questions, I made a radio piece for the MPR show I used to work for on the media. And the response I got was huge. Over a hundred people tracked down my contact info on the internet and wrote me. Most didn't even have a specific question or comment. They just seemed to want to talk about what was happening, about what it meant for them. A lot of them were musicians, but not all of them, programmers, designers, writers, people in completely different fields who heard the piece and thought, "That's my situation, too." And the more of these conversations I had, the more I realized something. Music might be the perfect way to understand what AI is doing to everything. It's older than every profession. It's older than every language. Everyone has a relationship to it. And yet, almost nobody was really looking at what AI was doing to music. So, I set out to discover what was going on. Over the past year and a half, I talked to over 40 people, musicians, CEOs of the AI music companies, industry insiders, lawyers, machine learning scientists. What I found? This transformation is simultaneously much scarier and much more alluring than I imagine. And also, I quickly realized this isn't just a story about music. To get a little grandiose, it's about what happens when machines start doing the thing that you thought made you human. So, where to start? But you might not realize it, the AI music revolution is already deeply underway. I believe that everybody's using it. You have Grammy award-winning producers, artists, songwriters, using these tools and their professional workflow. People are using it. It's very used in the industry. This was the overwhelming takeaway from the conversation I had. AI music isn't just coming. It's already here. But I probably shouldn't even say this, but when I talk to my artist friends who are serious in music, 100% of those people are adopting AI tools. And 100% of those people also don't talk about it publicly. I'm really all using it, but they don't talk about it. In reality, AI generated stuff as everywhere now. And a lot of people like it without knowing that it's AI generated. Those clips were from people who were here in future episodes. And there's nothing like a montage to make a point see more convincing. But ultimately, all of that was just a handful of anecdotes. It's not scientific. Of course, there's no data to base your judgment on. This is Helmets Benz, the CEO of Sonarworks, a music software company. A couple years ago, Helmets and his team realized a generative AI revolution was coming to music. And they wanted to get a slightly more scientific grasp on what was happening. So we said, "Okay, let's go out and talk to many people in many fields in the music industry." We ended up doing like over 100 interviews. To be clear, they were interested in generative AI. These more recent AI tools that generate entirely new sounds, whole songs. And what they found was surprising. There is definitely a debate out there in terms of is it good or bad? But of the people we interview, there was nobody who said, "This is crap and this is not impactful." I mean, everybody have tried it. They have realized that it actually works. It brings value and they have fully leaned into using it. Were you able to get a sense of where in the industry it's being adopted? I mean, I would assume it would be sort of on the fringes, like people trying to break in the industry. Is that where it was most prevalent? No, that's the interesting finding for me. The more you go on the real professional side, the actually faster they are to adapt. They are actually based with the economic forces of competition. You get paid by how much you actually work, how good and how fast you actually do your work. So the speed is a super-determinant factor. Once they realize that, "Hey, there is a tool that can help me in any way." They immediately lean in in a big way. Yeah, I mean, I guess that's why I was so shook by these tools when I started playing around with them. It was pretty clear that if I used them, I'd be able to produce more material, way better material, and do all of it a lot faster. I mean, something that can do all three of those things is pretty hard to ignore. The most surprising was the speed and scale of AI adoption in the industry. We assume most people are going to be conservative, very skeptical, not even considering the technology. And I was actually very surprised by the real scale and speed of adoption. And it's actually the professionals that are really, I think, driving it forward because they are fully in a competitive market and they view it as a competitive edge. And that's what really pushes it forward. The analogy that comes to mind is performance-enhancing drugs, like steroids in sports. Everyone could be opposed to it. But if you're a runner and you know everyone else around you is using steroids. Yeah, Mark, that's actually a brilliant analogy. And that's really how I see what's going in the industry. You know that steroids exist? You know that some people around are using them. And until the industry is sort of, unless it is regulated, I mean, you're kind of forced to also use steroids, otherwise you will not be competitive. Like, I think a lot of people, non-musicians who might be listening to this, might think, "I'm never going to listen to AI music, musicians I listen to, they're never going to use AI." If you agree with this premise, why does it matter if it is sort of working its way up from these people who are working musicians? It might not be household names. Well, actually, I'll start with giving you an anecdote. And unfortunately, because it's like sensitive data, I can't really name the names of these parties, but let's leave it as an anecdote. I was talking to people actually from a music label, and they were saying that, a very widespread phenomena is, the labels would have these contracts with artists right where, by this day, you actually have to put out a new album. So, this artist is out there, and they have created, let's say, five songs that they are happy about, but then they're really struggling with the other five. I was told that it's actually happening increasingly a lot, where the artists would just take generative AI and actually generate the remaining five songs. The interesting part was they were saying that they can't really fully control this. They actually don't have the capacity or technology to fully understand what is being submitted as actually generated content. So, you might be actually wrong in assuming that professional artists are never going to generate a song. Did the person say that those songs were noticeably worse or different? No, that's the point, actually. They can't even tell. We didn't go into how exactly and which tools they are using, but here's the reality. Like, there are already songs being submitted to labels where the streaming service or the label might actually not know how much of an AI has been involved in that. So, I think that reality might actually be coming quicker than we think at the moment. It might be here right now. And my answer to people who say that I will never listen to an AI song is you will never know. I spoke to Helmets in June of 2025. And over the past year, what he found with his survey has become increasingly true. A few months ago, Harvey Mason Jr., the CEO of the Recording Academy, you know, the Grammys, summed it up nicely on the Dakota podcast. I work in pop music, generally pop R&B, and in those genres of music, I think it's pretty omnipresent. When I'm in a room, AI is generally always there. It's being used to create chord progressions. It's being used to fill out drum loops. Some people are just creating entire tracks using AI. And besides the widespread usage, Helmets survey found something else, something I keep hearing from the people I talk to. They are saying, "I don't want to talk about it publicly." That was probably the second biggest surprise. None of them want to admit it publicly. So there is definitely a social stigma if you are a professional artist, and you go out and say, "I'm actually using these AI tools." You rightfully fear that you're going to be cancelled or something. You're going to have a lot of backlash because it's sort of unethical. Ultimately, I think this is why everyone thinks AI music is slop. When it's used tastefully by talented musicians, it's being done in secret. The audience has no idea AI is involved. Most of the people who admit to using it publicly just aren't making good music. So the only examples we have of AI music are bad, generic pop songs with horrible lyrics, joke songs about farts. And I get why talented musicians are using it secretly. People hate the idea of AI music. I do, using it feels like a betrayal, a polar opposite of authentic. And those who do admit to using it? Well, they find out why everyone else keeps it a secret. The summer, the rapper Tiger, said he used AI to make his new album. Pitchfork gave the album a zero out of ten. That's the first zero they've handed out in almost 20 years. Doja Cat gave it a shorter review on the live stream. Quote, "Tiger is a penis for making an AI album." So, the whole topic stays taboo, and that means this revolution is happening almost completely in secret. The reality is, everybody uses it right and left, but nobody's talking about it, and that's kind of the state that we are in right now. When you say everybody, you're kind of one of the only people in the world who can kind of back that out with some numbers, how much of everybody is it? 100%? Are we talking 80%, well, we talk to a hundred people from the industry, and the score was hundred. Like, everyone is using it in some capacity or another. Of course, we are exposed to a certain segment of a market, but I'm confident saying more than half of the professionals are actually using the tools already. But none of those musicians that helmets talk to thought that this was going to be good for their careers. And at the same time, they felt completely powerless to stop this infiltration of AI music. A very common comment that we got, again, anecdotal, is they were saying that, "You know what? Well, if this goes bad, I've done my sheer amount of work. I think I've contributed quite a lot to the industry, and I'm okay with that. So I mean, I'll probably retire when this guy now fully blows up." That was helmets, BEMs, CEO of SONERWORKS. Talking with helmets further rattled me. For starters, he showed that AI adoption in the music world was happening even faster than I thought. And I felt the same way as the composer Z interviewed. That when this blows up, I'll be out of a job, except I don't have the money to retire like all the industry vets he interviewed. But I kept thinking about one thing he said, that professionals are adopting the fastest because they have the most to lose. That's me, I'm the person he's describing. And here's the thing, 90% of the time I'm making music, it's for a client. It's a job. If these tools could help me do my job faster and better, don't I kind of owe it to my clients to use them? On one hand, the idea of using these tools felt awful. On the other, the only way forward professionally. It felt like an inflection point. Before I could decide what to do next, I needed to understand what these tools actually were, how they work, and what it means when a machine makes music. After the break, we go inside the black box. Welcome back. It's easy to get the basic gist of how these AI music platforms work. Udo, which we've discussed, and Suno, the biggest AI music platform, are pretty simple to use. You type in a prompt and hit enter, 10 seconds later, boom, music comes out. But what's going on inside this black box? It turns out it's actually pretty hard to understand. Because they're brains, they're brains. This is Dr. Maya Ackerman, a computer scientist known for her research in machine learning and music composition. And a core belief of hers is these systems aren't just similar to brains. They are brains. Sometimes people get mad at me. There was this hilarious exchange I had where a brilliant human being got mad at me for calling them brains because it humanizes them. Brains are not human. Brains are enrapt. They're literally neural networks. Right? What brain is a neural network? An artificial neural net has neurons that are artificial. We know they're not identical to our neurons, but it's a neural network. It's just an artificial one. I don't see any valid way to not call it a brain. And the thing about brains is that they're pretty complex. They're their own black box. Way above my pay grade. But what I can tell you about brains, you don't just program them, like a piece of software. People are used to thinking of computer science is something that the developer decides what the system does. But nobody just sat down and wrote Suno or Odio. Nobody programmed them to understand what a scale is, what a horn section should sound like, or how a Bernard Herman string arrangement builds tension. All you have to do with these systems, these brains, is feed them data. With Suno and Odio, it's estimated they were each fed 100 million songs. At that scale, Maya says something crazy starts to happen. The brain constructs itself. How can we not feed all of this? So, Alalams and music models can learn in a way that constructs their brain to be better and better and better as it more and more data comes in. These are self-growing brains. So, tell me if I'm thinking about this the right way. The training data, I think a lot of times I think of the training data as this is being put in as its memory. But really what the training data is doing is it's like building these networks. And so, like the real use of training on a shit ton of data is that it builds a realistic brain. Am I onto something or am I completely wrong? I'm so happy that you said that. I think a lot of people think that it's a memory bank. It's not a memory bank. Not a memory bank. I mean, if the listeners get one thing from this episode, that's the thing to get. Got it? So, how does it work? To vastly oversimplify, they just predict the next chunk of Odio over and over. And on enough data, the system just knows what sounds right. Which I have to admit is uncomfortably close to how I work. Here's the weird thing. As a human musician, the process of writing music is also a black box process. I'm never that conscious of what I'm doing. When I try and write a song, I get a vague idea or feeling and I mess around and discover the song. At a certain point, even if I'm super cognizant of what scale and mode I'm in and what the chords are doing, what's common in that genre of music, at a certain point, I have to just take a guess at what the next note is. Guess when the next instrument should come in. And guess when the next cool production thing should happen. The process for a neural network like Soono or Odio is basically the same. And that's a unique advantage AI has in music that it doesn't have in other fields. Writers know what they're writing, a designer knows what they're designing, the AI doesn't know what it's writing or designing and so it's just not as good as humans at those things. But musicians, we don't fully know what we're creating. So the process is much easier for AI to replicate. And there's another way that AI music has a huge advantage over other forms of AI. In music, there's no right answer. There aren't facts to get right, like when Chachi B.T. answers a question and there aren't rules of optics or human anatomy in the mimic, like when stable diffusion creates an image. People call the mistakes those systems make hallucinations, made up facts, extra fingers and so on. Music flips this on its head. I heard you on a podcast and the guy interviewing you said something like hallucinations aren't all that bad in music, but I was thinking about it and I realized that's a big understatement. Making music is a hallucination. It's not something that's just tolerated. That's what making music is. Exactly, that's exactly what it is. Now of course, when somebody is being imaginative, it's nice to know that they're being imaginative. So what Elon's do when they present something as fact when it's actually hallucinating, I can see why that's a challenge, right? But if we simply understand what's happening, if we recognize that these machines are prone to imagination, there isn't this false expectation. Just let the machines imagine and when you're in a creative domain, it's exactly what you said Mark. It's the complete hallucination and that's what it's supposed to be and it's beautiful. But here's what I keep going back to. If it's really working like a human brain, why does something feel off with a lot of these AI tracks? Take this track I found that someone made and posted on UDO. The description made clear that they were trying to make a song that sounded like Radiohead. And it does sound quite a bit like Radiohead's album "Kide" without ripping off one particular track. But the voice? sounds way too much like Radiohead singer Tom York. A human musician, influenced by Radiohead, would never make something that sounds this much like him. I've tried to make a Radiohead song before, and even if the starting point was basically a prompt of make a Radiohead song, the end result at most would just sound inspired by them. ♪ No you don't think ♪ ♪ About ♪ I've never played this song for another person, but it's not good as it is. It's still better than that Radiohead rip off from audio. Because my failed attempt to sound like Radiohead at least became something new. And that's how music works. I think this is a really key point. Take the Beatles. They were trying to sound just like Buddy Holly or Elvis, and their failed attempt is what became the Beatles. That's how music evolves. But AI doesn't fail in the same way. Instead, you get a pretty soulless track with a creepy Tom York clone singing. It's a business choice. They know it's easy to cheat around, but if they want it to, they can make it uncheatable. They don't want to. There is absolutely no excuse for these models to imitate anybody. No excuse. You can learn from data without ever imitating anybody. Soono and Udo say these are mistakes. Just like when human musicians accidentally write songs that happen to sound like previously written tunes. I can't say definitively whether it's a feature or a bug, but it does feel different than when, say, Sam Smith writes a track that sounds a little bit like a Tom Petty song. Take this example. I wanted to create a track that sounded like Glenn Gould's Goldberg Variations. So I prompted Udo with just eight words. Bach, solo piano, virtuosic, theme and variation, 1956 recording. You could say it was just following a prompt like a human composer would have, but this is different. Listen closely. You can hear Gould's very idiosyncratic moaning in the recording. I'll isolate it the best I can't. Hear it? Here's a real Glenn Gould recording. Hear that moaning? That sound, Gould's personal fingerprint, feels like proof to me that Udo is mimicking a specific recording. In other words, copying. And even if it's not copying, it sure feels like stealing. Take this track that sounds like a young Aretha Franklin. Or maybe it's more like Eda James. Either way, it's clearly mimicking that late 60s sound of famed studios in Alabama. And most importantly, it has created this voice by ingesting tens of thousands of gospel performances. And that's no longer just about copyrighted recordings. That's about co-opting people's spiritual practice in exploited oppressed group spiritual practice. It feels like stealing their soul. It feels sacrilegious. And this is why so many people find AI music repulsive. That's why I didn't wanna use it. And yet, whether I used it or not, whether it's a moral, it's still coming for people like me. It's coming for everyone who makes a living, doing what these machines can now do in seconds. - On a podcast I heard you on, you said something to the effect of, well, background music or music in a film where you don't really care about it. That, yeah, might get replaced. And, you know, as I heard you say, I was like, hey, that's my job. That's my livelihood. - Oh, I'm sorry, did I say it now? - You might not have said it that harsh. - I think the truth is that I think that this is gonna get cut into. There has been efforts to cut into that for a long time now. And I think they're gonna be more successful than some other intrusions into human music making. Now, that doesn't mean that it's okay. It doesn't mean that it's painless. It's just where there's a lot of effort going. Unfortunately, it looks like there is gonna be some success, right? Just because I say something's gonna happen doesn't mean that I think it's good. Of course, the obvious move here to save my job would be to just use these tools myself. Fight fire with fire. But I felt like the worst thing I could possibly do. And then something happened. After the break, what changed? So I was ready to write the whole AI thing off. I had poked around just to see what they could do. And now it was time to go back to making music the old way, the way I knew. And I would have, except I made a mistake. Just before deleting my account, a little window popped up promoting a new feature. And I tried it. This feature on UDO, it allowed you to upload, you know what I mean? It allowed you to upload your own music. The idea was that it could listen to what you were doing and just continue it. To test it out, I uploaded a 12-second jingle I made a decade ago for a commercial that I pitched and didn't get. Nothing special, but I always thought that could have been the start of something. So I uploaded it and with a little prompting, it turned into this. It starts exactly the same. But then, completely seamlessly, it keeps going. This was more mind-blowing than the tracks I heard it create from scratch because, well, it doesn't feel like some other musician. It feels like me. And that felt different. It wasn't conjuring Aretha's ghost or channeling Tom York. It was extending something I started. The soul and the machine, if there was one, was mine. Take this drum machine. I totally had a vintage CR-78 in mind, and that's exactly what it added. I started imagining all the things I'd always wanted to do with my music, but just didn't have the time. Or the budget. This was an old guitar demo, I found. I don't even remember making it. But listening to it, I thought, huh. This could actually sound cool with a good horn section. The problem is getting real, tasteful horns on a track is expensive. I've always wanted them, just never been able to afford them. But now, I uploaded this demo to UDO, prompted it, and it came up with this. It's switching to the AI version right now. Then, Soono came out with something even more ambitious. While UDO had been extending my ideas or putting something on top of them, Soono offered to transform them. You could upload a rough, rough demo, and it would spit it back as a fully produced track, vocals and all. So I found this super old, super rough, very embarrassing demo to test it out. I uploaded it, typed in some lyrics to replace the bump, bump, buzz, added some genres like psychedelic pop, indie pop, and out came this. It's just something so powerful about being able to have an idea. Record a really rough version of it, and then 10 seconds later, get more or less the fully produced version you're hearing in your head. This really felt like a tool that could help me make my music. It felt less like theft and more like empowerment. It felt magical. You have a vision in your head. You can realize it instantly with this stuff. And like you said, it aligns with exactly what you wanted. It's kind of like, if I may, God said, and it was there. With Gen AI, don't write. We imagine something and then it's there, right? It's an ability to create at a speed and an agility at a power that's just unprecedented. Hearing Suno turn old songs, I had completely abandoned. Just left to die on a forgotten hard drive. Hearing it turn that into something cool, really cool. I could see what Maya was saying. It felt like having an amazing bandmate who knew my musical sensibility intimately. a bandmate that could help pull the absolute best possible music out of me. That's how AI makes a human being more creative. I think we have to admit that there is a new kind of musician, and that musician is AI, and we can work with it. I didn't want to admit it, but I was excited. And it wasn't just that I might be able to keep my job if I started using these AI steroids. I have tons of unfinished songs. Summer produced tracks I couldn't quite finish. Summer of 15 second jingles I did for commercials that didn't get made. Summer voice memos. With all of them, I didn't finish them because I got stuck. I think every musician, maybe every creative person, can relate to this. It's so easy to come up with the germ of something cool, but it's so hard to get it from 70% done to 100% finished. Now, I had no excuse for leaving a song unfinished. And it wouldn't even necessarily require using the audio from Suno or Udo. I could use these AI music systems just to come up with ideas to get unstuck. Let's say I can't figure out a chorus for a song. I could have Suno generate 10 different choruses. Take the chords from one, the bass line from another, perform it all myself, and end up with something that doesn't sound like any of them. That feels way more like spitballing with a writing team rather than stealing someone's musical soul. But I was still conflicted. Very conflicted. The morality, the legality, the qualitative changes that just inherently happen when your collaborator is AI. But even just playing around with them, whether I ended up using them or not, it was shocking to see how easily AI could do something that up until a year ago, very few people could do. Make a song. And that was part of my identity. It was my thing. You could give me a prompt or a commercial or a film, and I could make a song for it. It was the thing that made me a little special, and I suspect most people feel this way. Being a good lawyer, copy editor, accountant, even if those aren't the flashiest jobs in the world, being good at them can still be part of your identity, what makes you you. What happens now if a machine can do it faster and better? I asked Maya about this. I've been reaching out to a lot of my musician friends and showing them how all this works. When they see it, they're pretty blown away, but I think they have a lot of the same conflicting feelings I have. But one takeaway for a lot of people, including me, is it shows how what we're doing as musicians might not be quite as creative as we thought it was. It kind of shows how formulaic it can all be. Oh, I'm sorry, sometimes I have strong opinions. Why do we immediately declare something valueless if computers can do it? Why is it not creative because a computer can do it? Well, I guess it's just like, if something can do it so quickly, so painlessly, why should I do this? And then it's like, oh, that thing, it used to make me special, and now anyone can do it. Now, that is the truth, but you shared right there. That is the pain for us in this transformation. It doesn't matter if it's creative or not. What matters is that it used to make me special, and now it doesn't. I think it's really interesting, Mark, that it took us, we had to go through so many layers before getting to this. Because now we do have to grapple with the fact that something that used to be special about our abilities is not special anymore, and so we need to find where are we special again, and we're going to do it somehow. Maybe I want a believer, but the overwhelming feeling for me is loss. It's not the AI stole from any particular musician, even though it definitely did, and it's not that it will eliminate musicians' jobs, even though it definitely will as well. It's that this thing that used to be magical, dreaming up a song and a thin air, has become banal. On a societal level, it was a special piece, a core piece of what made humans human. And if it's not just ours anymore, what does that say? There's always been something special about music, and maybe part of it is that musicians can't fully describe how they make music. It's always been a black box. For me, for every musician I know, you don't just decide what the next notice. You just guess. And now seeing a machine do that so easily, it makes me wonder, what if this whole time, what seemed like inspiration in human musicians, in me, was actually just probability and pattern recognition? That question is not just about music. It's the question AI is forcing on all of us. What parts of our work are actually us? And what parts are just patterns we've learned? And if a machine can replicate your judgment, your taste, your instinct, was it ever really yours to begin with? Some fields can dodge that question for now. They can say, "Well, the AI isn't good enough yet, and they might be right for a year or two." But in music, it's good enough now. That's why so many musicians are secretly using it. This means musicians are some of the first people who actually have to sit with these very difficult questions. There's no escape hatch. There's no give it 10 years. So over the season, we'll use music as an entryway into these thorny questions that all fields, all people, will soon need to navigate. So there are definitely hit songs that are out that have been made with AI. I know that. We're going to talk to the hit makers, the people who write the songs you hear on the radio. It's essentially a plagiarism machine. We'll hear from musicians fighting back, making the case that this whole thing is theft, plain and simple. It takes our work and transforms it into something that it claims we don't own. We'll explore what happens when the tools get shut down, when people lose something they've come to depend on. You know, I guess it's like the stages of grief in a way. It was like the perfect tool. It was like having an orchestra in the house. I was not in a great state, and idiot came at the right time. Suddenly found there was a way of expressing myself. And we'll go to the Suno headquarters to ask the CEO, how do you justify training on millions of songs without permission? What we've done is not illegal, and there's no violation. The lower end of the damages is in the billions, and the upper end of the damages is in the trillions. This isn't just about music. It's about creativity, ownership, what we automate, and what we keep as our own. It's coming up this season on Lightbox. Lightbox is reported, written and produced by me, Mark Henry Phillips at Windhill Studios. The place for narrative podcast production. The producer is Drew Thurlow, and our assistant editor is Ryan Seaton. Big thanks to Helmitz Bems at Sonarworks and Maya Ackerman, author of the book Creative Machines and Founder of the Company Wave AI. So we make this show completely independently because we love these kinds of shows and these types of stories. If you do too, please support us right now by rating and reviewing this show. It makes a huge difference and can really help make or break this project. Do it. Like, right now. Finally, Lightbox is a proud member of Radiotopia from PRX, a network of independent creator-owned listener-supported podcasts. Discover audio with Vision at Radiotopia.fm.