Taking Eight Million Creators to Eight Billion People

Richard Kramer:

Welcome back to Bubble Trouble, conversations between two humans with real pulses as opposed to computer generated voices. It's the real life double act of myself, independent analyst, Richard Kramer, and the economist and author, Will Page. And this is what we do, lay out the inconvenient truths about how business and financial markets really work. We're now deep into our inquisition about AI and ChatGPT giving you dinner party conversation fodder to make you appear smarter for the next six months. And we're going to get another layer of training data today, talking to some of the biggest names in AI.

We want to get into the controversy over Boomy with the co-founder, Alex Mitchell. To date, artists, and I say that in inverted commas, on Boomy have created, with the help of some clever algorithms, 16 million original tracks. That's more music than iTunes delivered at its peak. And most of them, apparently, are now available on Spotify to stream after some controversial take down episodes. Is this AI in action, a dystopian future or the shape of things to come? We'll be back in a moment with Alex Mitchell, co-founder of Boomy.

Will Page:

Alex, welcome to Bubble Trouble. It's a real honor to get you on the show. Things are moving fast in this space, which makes this podcast timely. And that 16 million tracks, I calculated that at the very peak of iTunes, that's how many songs they had to offer. Your company is offering the same volume as iTunes at its peak. The key to kicking off a great podcast is to assume no prior knowledge, no prior knowledge at all. And let's assume that, and you tell our audience what Boomy is. Imagine it's a kid in a bedroom putting songs up on DistroKid or a parent of a kid in a bedroom putting songs up on DistroKid. What do they need to know about Boomy?

Alex Mitchell:

Yeah, sure. At Boomy, we have a pretty simple mission, which is to transform ordinary humans into creative musicians. And you mentioned the 16 million songs. One of the things that I would just start out with here is that we see that as a small number, not a big number. When you think about the musicality of humans in the world, how many musicians are there? Probably somewhere between eight million?

Will Page:

I'd put it at eight million.

Alex Mitchell:

How many million? Eight? Eight to 10 million?

Will Page:

Spotify say there's eight million creators. Yeah.

Alex Mitchell:

Yeah. There are, what? Eight billion humans? Something like that, I think?

Will Page:

Interesting.

Alex Mitchell:

To some degree, I really believe that every person, every human being, is musical. As musicians, and as a musician myself, what do we do? We do two things. We create a song and then we listen back to that song and we decide what we should do with it and how we should change it. And that tends to be tied to the education we've received, the specialization we've received.

I was very lucky to have a public school that had great music education and come up in a family that was musical, but most people in the world, and particularly in emerging markets, don't have access to the time, the training, the education, the tools needed to express themselves with music. And so AI, to me, is a function that we can use to draw musicality out of every person on the planet.

And I think one of the analogies that I like to use is photos. We've got, you're saying, 16 million songs, iTunes at its peak, there are 400 million, 500 million photos, and God knows how many videos, that go to social media sharing networks every day, every 12 to 14 hours. The difference being specialization and the desire for people to share visual imagery and communicate through visual imagery.

And what is really exciting, to me, about the potential for this technology is to draw that musicality out of people. Again, I think a lot of people want to talk about AI and the technology, and that's super interesting, but fundamentally, what we're here to do is if you want to express yourself with music, we want to give you the tools to be able to do that regardless of your prior specialization.

Will Page:

We're off to a flyer and let's just build that point out. Eight million out of eight billion is a fascinating point to think about. It really is. I was on the BBC show with Samira Ahmed, a show called Front Row, and we had this performer called Lisa O'Neill in the studio. And she said that we were exposed to our first music, our first rhythm, before we were born. She's quite adamant by this point. Those eight billion people have all been exposed to rhythm or are interested in rhythm, and you're opening the floodgates to get them all involved.

Again, to reiterate what Richard said in the introduction, the key to a good podcast is not to have a great conversation, but it's to make your listeners look good at dinner parties. I've got a couple of higher bound dinner party questions for you here.

Alex Mitchell:

Sure.

Will Page:

And we take AI as a catchall term. Do you think AI, as it currently stands, is a problem in search for a solution, or a solution in search for a problem? What is the actual use case?

Alex Mitchell:

Yeah. That's a good question.

Will Page:

You've illustrated one, but that higher bound, what are we actually doing here? Before we get into the rabbit hole of what you're up to, let's just dig up the upper bound there.

Alex Mitchell:

Yeah, look, I think one of the big challenges that I've seen in many of the conversations that I've been having with a lot of different people in the industry over the last couple of months, is this very loose and very fast changing definition of what we mean when we say AI.

And so in order to look at the question of, well, is it a catchall phrase to describe a product? What is AI? And to me, my own definition is, well, it's really just a shorthand for automation. That tends to be how it's used in practice.

I'll give you a great example. I don't think that anybody that I know in the AI research field, anybody on the music side, really anybody would call a otherwise normally produced song, in a studio, usually a digital audio workstation, using traditional methods that happens to have a deep fake vocal on it as a AI generated song. The notion that you would call that AI music is less than six months old in my view.

Will Page:

Right.

Alex Mitchell:

Which started with sort of the fake Drake stuff and all that, saying, "Wow, look at this AI generated music. What are we going to do?" And you're like, "Okay, there are Elvis impersonators. Impersonation isn't dependent on AI." And if you're looking at that song and calling it AI generated, that is a very different definition than what I would say is AI generated.

I think it starts from just saying that this is a fundamental set of technologies, the methodologies of which are changing very rapidly, are advancing very rapidly. And I try to use the word automation as much as possible. I try to use other terminology as much as possible. Because I agree with you, it can be a confusing catchall term that people jump to the end of the world with very quickly sometimes.

Will Page:

Let's move from that to the hysteria that surrounds AI and music. You've already touched on a couple of interesting points with the Elvis impersonator example. This week I think I've been asked to keynote four conferences all under the title Is AI Killing Musical Creativity?

Alex Mitchell:

Is it?

Will Page:

For God's sake, man. This hysteria has got to stop. Please. And I want you to stop it. I want you to stop it now.

Alex Mitchell:

Yeah.

Will Page:

To tee this up, I want to get historical context. When we went through lockdown, I had a terrible lockdown, I set myself the goal of three things. Learn how to run, I have to thank my co-presenter Richard Kramer for getting my 10K under an hour, learn Spanish, I failed abysmally, and learn how to play Def Leppard's Hysteria on guitar. And I was shocked to learn that Joe Elliott had a computer manipulated voice during his career. He never got up that high. Computers got him up that high.

Can you just cool the heads for a second and give us a calmer interpretation of what AI is bringing to music compared to the the wolves are at the door hysteria that we're hearing in the press? We've been manipulating computer music for years.

Alex Mitchell:

Exactly. Let's kind of dive into where the hysteria is coming from then. If there's a few things that are obvious, it's obvious that people have been using algorithms to create music for a very long time, and rules-based approaches even before that. It's obvious that some of the music and artists that we enjoy have been using automation or even various forms of AI for a long time. I was talking to a very well known producer this morning who's like, "I've been using AI since 2014, 2015. It's in my digital audio workstation." [inaudible 00:08:21]. Right. There have been these automated processes for a long time. Why is there hysteria? Where does it come from?

And I think in order to have hysteria, you have to have a fear. People fear what they don't understand, and I think people fear what they can't control. And I think that there's a lot of noise and a lot of lack of understanding, but let's be clear about what we are afraid of, we meaning the music industry, is afraid of here. I think that a technology in and of itself existing is not something to be afraid of. And that feels obvious. If you had some sort of doomsday weapon, but nobody could press the button on it, it wouldn't be scary. It's only scary if it can cause harm.

I think what people, this is just me editorializing here, for the dinner party, but I think that what is scary is actually those other eight billion people, that other 99.9%. This idea that you're going to take this sort of inevitability, because we've been trending this way for a long time, the number of artists. Automation has been creating more artists for decades and decades. It's been simpler and simpler not just to create, but also to distribute, to your point, and to participate in the music economy. And eventually, if some of the principles I laid out in the beginning are true, which is that most people are musical, most people probably could create music or would want to express themselves musically if they had the tools to do so. Well, eventually, that's going to converge to a point where the tools are so simple and the interface is so easy, that interface now probably being text and natural language, that you're going to have a lot more participation.

Are we afraid of AI or are we afraid of greater participation in this way that was maybe happening faster or accelerating an existing trend? And I think that's where you get into the interesting conversations about the impacts of what's going to happen in this universe where you've got maybe even a billion artists, a billion musicians, eight billion musicians. That's the really kind of interesting sticky stuff I hear that I think people are having lots of talks about.

Will Page:

I hear it.

Richard Kramer:

I want to push back here on something you said because I think it's a very benign and it's a techno-optimist, techno-futurist kind of reversion to say, "It's just tools. We're just democratizing this medium for everyone."

Now, the simple economic reality is that AI has been used to create an enormous volume of production music, which has flooded streaming platforms and has diluted or diverted some of the funding that would go to legitimate artists practicing the craft as we knew it. And maybe you're saying we should just burn down this previous notion we had of the craft as we knew it, playing instruments, for example, and enter the grand new world, but you can not dismiss that that democratization of tools has been a net negative for a lot of legitimate artists who followed the old school path.

Now, are you telling us we just need to jump that shark? Are you telling us there's been none of that pernicious economic impact, which I think we would argue against, but I'm struggling to accept this pure autonomous technology tools based vision that just says, "Oh, we're going to democratize a billion or two billion Bachs or Beethovens or Princes or David Bowies."

Alex Mitchell:

Yeah, sure. Look, there's a few things in there that we can get into. I would actually take it one step at a time.

In no way is that my argument, to be honest with you. I think that saying it's just a tool is a coping mechanism for the fear that I just sort of laid out. It's saying, "Well, we don't have to be scared of this thing." And I'm not even saying that there aren't some legitimate concerns, but I'm going to take issue with... I guess describe to me a legitimate artist, because I think the narrative that there's all of this AI generated music out there. I would just take issue with the premise. I've always taken issue with this premise over the last few years. There is, at least with the way that we operate, there's a heartbeat behind every single one of these songs. I've never met an AGI that decided in its AI wisdom-

Richard Kramer:

Infinite wisdom.

Alex Mitchell:

Yeah. That the thing that it decided to do, it said, "Hi, I'm an autonomous AI, I think for myself, and the thing that I want to do is produce a song." No, it's that these are people, these are human beings who are using... I don't know. Just a tool feels wrong because it is so powerful, what this stuff can do, but are still using technology to express themselves with music.

And I'm going to take an issue also with the notion of dilution. I think if you're pouring water into water, you just have more water. That's not dilution. And I think that what you're getting to here, when we talk about there's such a thing as a legitimate artist and not a legitimate artist, that's a debate that I don't want it to minimize it because I think, look, I'm a violinist. Okay?

Richard Kramer:

Right.

Alex Mitchell:

I think you can draw a purity line anywhere.

Richard Kramer:

Right. A speeded-up version of someone's track published by a different artist, a speeded-up version of a track, is that a legitimate artist or is it theft?

Alex Mitchell:

Well, that feels like theft to me, but that isn't anything close to what we're working on. I think there are some apps that do that, sure.

Richard Kramer:

Yeah, there are, and there are many companies who've monetized a business model that says, "We will just make subtle variations or white noise tracks or what have you." Using automation, they will generate random music and get it inserted into playlists to extract some of that pool of revenue that's attached to the entire music industry. And that's been a net negative for a lot of people who wouldn't take that approach or would use automation more sparingly, not as the core of their business model.

Alex Mitchell:

And I think if the designers of those apps, and I think if the people who are working broadly in the AI music field were responsible for creating the economic model that you're talking about, I would understand what you're saying, but I think that part of the conversation here, and I think, again, does this actually have to do with AI or does it have to do with lots of new creators coming in and this notion that a 32nd white noise song, which we don't do at Boomy, but that is a thing that happens, should that be paid in the same way? That is up to the platform. That's up to the consumer. That's up to the market.

Wrapped up into all these conversations is this conversation about a change to the streaming model, which I've got to say I could not agree with more. The notion of the pro rata as a musician, as an artist, and as an independent artist never really made a lot of sense to me, to be honest with you. And it opens up, I think, the door for some of what you're talking about, but ultimately the market's going to decide how those rules change and if there needs to be things like content standards.

If we could produce a billion songs next year, are all of those going to every DSP as it exists today? I don't know. We're midpoint or probably early in figuring out what this new world looks like. I would reject the notion of techno-utopian. I really think that we're approaching this as a team of artists and as a team of musicians and as people who really do deeply care about the... let's call it the first 100 years of recorded music. And that's part of why we've been really loud about advocating for what does it look like for payment models? What does it look like to pay artists for training, to pay rights holders and publishers and everybody in the ecosystem? If it's true and kind of somewhat inevitable, AI or not, that there's going to be billions of artists in the future, there are systems that we're all very familiar with that need to be updated.

Richard Kramer:

Yeah.

Alex Mitchell:

And maybe the pro rata model is one of those, that's not me, how that's paid.

Will Page:

Well, we can go down that pro rata rabbit hole in part two. That's something that's close to my heart, but I think there's a few things I want to tease out here. One, I'm very interested to learn of the stickiness of AI music. You're supplying a lot of content, but are you actually seeing demand, momentum, these types of compositions moving up the charts? For example, backstory, 20 million people streamed a fake Drake track. 20 million once each, nobody streamed it a second time. And that's what I found interesting there, but are you seeing genuine momentum happen with this AI music? Are you developing fan bases?

Alex Mitchell:

It's a good question, and I think of it very differently, and I think we've always thought about this differently, in that sort of on one side of the table, we're having a lot of conversations with industry and we're having a lot of conversations along those lines. Can we fit this or can we think about this in terms of are you going to produce a hit? Are you going to produce a charting artist? On the other side of the table, we see our user base, which of course is growing like crazy, where you have these people who are so inspiring, who are creating music oftentimes for the first time, oftentimes for highly personal reasons and personal context. You're talking about parents and kids making songs about their bedtime routines.

Will Page:

Wow, music for the purpose of moment.

Alex Mitchell:

You have educators and teachers are using it inside of-

Well, I'm just saying there's so many just hyper-personalized contexts that people start creating music in that is a totally different, and frankly, very exciting and very new opportunity in how to create meaning. Because if you'll let me rant a little bit longer on this point, I think this is so important, that what the music industry does in a simple way is we try to create meaning around music. The more meaningful a song is to somebody, the more likely they are to continue to consume it and it drives economic activity.

Will Page:

Intimacy.

Alex Mitchell:

You can do that at a show. Intimacy, yeah, you can do that at a show, live show is a great way to do that. You can do that through a marketing campaign. You can do that through all the different things that this industry has figured out how to do. Or get it on a chart. "Hey, it's this chart position or that chart position," that imbues this song with a lot of meaning.

And I think what we've hint on here is that there's another way of creating meaning, and I call this your five favorite photos problem. If you think about your five favorite photos, you're probably in all of them, or somebody took them of you. And it probably doesn't belong in MoMA. It's maybe not the greatest photo in the world, but it has a meaning.

I do think it's inevitable that at some point, yeah, you're going to have songs that start to do pretty well for a wider audience here, but there's so many parallels, I think, in video mediums, in what we're seeing on social networks, where if you can tap someone's personal connection, that's what we're seeing drive consumption. And we're seeing that consumption drive in a different way than I think the industry is used to thinking about music. And that's what's exciting for us and where we want to orient as a platform.

Will Page:

I hear you. I hear you.

Now, before we get to the break, just a couple of points to wrap up on. Alex, if it helps put wind in your sail, I can remind you that in the 1980s when dance music got going, thanks to the godfather of DJing here, Greg Wilson, friend of the show, session musicians tried to block dance music from Radio One because they felt it would put them out of business. We've been here before. We'll be here again. Or as Mark Twain said, "History doesn't repeat itself, but it sure as hell rhymes."

Alex Mitchell:

I think there is so much history here. There's nearer history, there's far history. If you go back, somebody sent me... There's these political cartoons and they're from... I want to say the early 1920s, don't quote me on that. It was the advent of recorded music, and it was, "We're going to put recorded music in theaters."

Will Page:

Yeah.

Alex Mitchell:

"And you don't need a pianist anymore, you don't need a musician anymore." And the musicians union built these... they created these cartoons of these evil robots with this big sack of prophets coming for... The concept of recorded music itself at all has been here and was fought and had. There's so much historical precedent, to your point, with synthesizers that were pushed back on, with auto-tune that was pushed back on, even drum samples to some degree had some controversy around it. And so to me, it means that we are going to go through that cycle. I think we're ready and prepared for that cycle. It's going to be very in vogue to, I think, beat up on AI media for a while, and then we're going to blink in five or six years, it's going to become very normal. That's my prediction anyway. I could be wrong.

Will Page:

Well, in part one, let me just wrap up in part one there real quick because I want to get into the training data debate in part two, but I said earlier on, I learned Hysteria and guitar. I think what you've inspired me to do is to go to Boomy, recreate Hysteria and ask ChatGPT to write the lyrics to calm peoples down about the nerves they have around AI music today and see where that goes. We could even get Joel Elliott to sing in his true Voice and not have it computer manipulated like it was in 1987. But listen, we'll be back in part two. We are going to go down a rabbit hole on your training data. We'll be back in a moment. Thank you so much.

Welcome back to part two of Bubble Trouble, myself, Richard Kramer, and Alex Mitchell from the AI music company, Boomy, which has on its own put up more music than was available at the very peak of iTunes, over 16 million songs from this relatively speaking unknown company. We're going to know a little bit more about it. And Alex, in part two, we would like to go down the rabbit hole of deeper topics that we've maybe skimmed the stones of in part one.

I thought the first one I've got to get to is the training data question.

Alex Mitchell:

Yeah, sure.

Will Page:

Now for people who don't understand AI, training data is such an intuitive term. You've got to give the machine something to learn from before it can perform its magic, and that's where humans design algorithms too. But your training data, from my understanding, is different in that you're not requiring copyright music to get the Boomy effect to boom.

Alex Mitchell:

Yeah.

Will Page:

Is that a fair description of how it works and what makes you distinct?

Alex Mitchell:

Yeah, yeah, absolutely. And this goes to what I was saying before about how AI is this catchall term. There's a lot of different ways to create a song. I think the first waves of AI music, companies were not using some of the training mechanisms that people are talking about today as it pertains to LLMs. And we've been around for a while. I think what's core to understand about our approach is that Boomy is a team of musicians. Again, we care about this industry, we respect copyright. I would hope that our copyrights have the same sort of protections that anybody else's would. And the notion of taking someone's work or taking the work of a songwriter, training a model with it and saying, "Sorry, that's mine now." That does not sit right with me as an artist. It doesn't sit right with me as a technologist. And I think if you're going to do that, there needs to be a way to pay for it.

But in terms of how we work today, look, there have been algorithmic musicians and ways of producing music in particular styles with algorithms for a very long time. And the best way I can describe this is that we're more on the Brian Eno spectrum to start where you're creating algorithms that can produce in styles that are maybe not quite as accurate as you would get to if you were training a model on data that was subject to copyright. But over time, especially through our user base, we can get to a point where, to the extent that we're training models, we're only doing it on outputs of algorithms that we have internally developed and that we have internally refined as artists. You can think of this kind of as being any other algorithmic musician in the pre AI era. And we're just sharing that with other people.

You can think of if somebody created a music algorithm and then invited somebody into the studio to manipulate it and use it, that's how we work. That's how we've worked for a long time. I do think there are incredibly exciting technology advancements and massive opportunities for this industry in the licensing of training data. And that's very much where we want to go and very much the conversation we've been in.

Will Page:

Just one more on that one there just real quick. Just so our audience is clear, if parents learn that their kids are using Boomy, they know that they're not infringing copyright material because Boomy itself is not training off copyright material.

Alex Mitchell:

Yep.

Will Page:

That's absolutely crucial to understanding what you guys are doing. I think there's a lot of misunderstanding in the press there. And that means, if we move from one acronym to another, you're pretty much the closest thing we've got to AGI. Music is always the first to suffer and the first to recover, but you're not actually needing that human element to get going. Is that fair? Are we leaning into the AGI space with what Boomy's doing?

Alex Mitchell:

I don't know that I would take it that far. I would say that you're looking at...

Will Page:

Maybe peering into the window of what AGI might look like.

Alex Mitchell:

Look, no, I think we have a mission, we have an objective, and that's to draw musicality out of people. We're going to use the best technology that we can use and the best systems that we can build in order to achieve that goal. And look, years ago as we were scoping this and there were researchers and people and a lot of these models that were based on the use of that copyright training data, it was kind of like a, "Yeah, someday that's going to become a problem. Let's try to figure out a way where we can get started, draw creativity out of people."

Will Page:

Circumvent.

Alex Mitchell:

Not circumvent, but get to a place where there can be a really positive and active conversation about licensing training data. And I wish that conversation happened three years ago. That conversation is more happening now. I think ChatGPT really changed the game in showing everybody what can happen, at least with text or with the other media domains. And as we prepare for music's ChatGPT moment, so to speak, I do think there are, again, it's one of the biggest opportunities you could possibly fathom, that if there are going to be tons and tons of songs in the future that are coming out of these systems and you have a way of paying the artists and the songwriters and the rights holders inside of those systems, just imagine the universe of revenue that that can be created. And so that's what I just think is super interesting and why we've been so active on that front lately.

Will Page:

Richard.

Richard Kramer:

Again, I'm going to take the skeptical position here.

Will Page:

Sure.

Richard Kramer:

Because I'm going to go to the godfather of algorithms. And you're playing with words, you're giving us a sort of subtle evasion. Because Bach, let's call Bach the godfather of algorithmic music.

Alex Mitchell:

Sure, yeah.

Richard Kramer:

He's a mathematical genius. I don't know if you saw the film Tár. There was a beautiful scene there where the Cate Blanchett character is defending the beauty of Bach.

Alex Mitchell:

Of course.

Richard Kramer:

The essential beauty. And he's over 70 years old, gone over 70 years, so he's certainly out of copyright. Now, does Boomy use Bach to train its algorithms? And if so, do we want to freely take from the rich legacy of musical history just based on what's available under a legal definition, really that pretty much came out of one country?

Alex Mitchell:

Sorry, so your question is? We don't train any models on Bach today. No.

Richard Kramer:

He's not in copyright. Some other AI could come along and say, "It's not copyright music. We don't have to remunerate anyone. We can just take one of the most brilliant classical musicians," and really a godfather of algorithmic music, one could think of him that way, and use that as your training set?

Alex Mitchell:

You could also take Bach and you could give it to a really novice performer and have them perform it in a horrible way. And how do you think Bach feels about that?

Richard Kramer:

He's long gone.

Alex Mitchell:

Right. I think I understand why you're asking this, but I would say that if there are people out there who want to appreciate and who want to express themselves... If you're a musician, let me simplify this. If you're a musician, and I've played plenty of Bach over the years as a violin player.

Richard Kramer:

I'm sure, as a violinist.

Alex Mitchell:

You study it, you learn it, you appreciate it, and you have time, you have time to do that, and you would be lucky to have time to do that. Again, we come at everything through this lens, and I'm not trying to be evasive at all when I say it, this is what we're trying to do is bring musicality out of any human being in the way that they could if they had that time and that education and everything that I was lucky to have. Not everybody can take violin lessons. That doesn't mean they shouldn't be able to express themselves with the sound of a violin and with the influence of those great composers.

I think embedded in your question is that there's some sort of problem, that we've take in this thing that people are doing, which is listening to Bach, becoming inspired, expressing themselves with a lot of automation, you could drag the MIDI file of a Bach song, put it into a digital audio workstation, and you could put some synthesizers on it and there you go. I guess why is training a model with Bach problematic if that other thing that I just described, which is taking that same composition and either creating your own version of it or creating a really bad version of it, what is different to you? It sounds like you're saying that one of those things is okay and one of those things isn't okay, I guess why is that, especially as it pertains our topic?

Richard Kramer:

I guess what I'm trying to get at is do you protect the kernel of invention that stands behind 500 years of music history, or do you say that once it's older than 70 years, it's all fair game? And once you do that, you open up a whole range of music that can be expropriated without remuneration? Because I'm just pushing back on this notion that you're going to create a universe of new creator income, and it's built on the backs of people. "Okay, well, it's 70 years gone now, so maybe their heirs are... they're out of luck." And this is a wider point about AI because it's about standing on the shoulders of those who came before you.

Alex Mitchell:

Yeah.

Richard Kramer:

It's about the difference between organic creation and borrowing. And this very wooly line between plagiarism and invention.

Alex Mitchell:

Yeah, sure. Yeah, I think I understand your question and I don't know that I'm the right person to answer the copyright policy on the 70 years.

Richard Kramer:

Yeah.

Alex Mitchell:

It would also suggest that you could create a streaming service or some other sort of use of this content after that expiration window. Not a question I've gotten before. It's definitely something I want to think about. I don't know that I'm the best person to talk about expiration windows on copyright. I think where I can be helpful in understanding this stuff, and I think the understanding can reduce some of the fear, is just talking about how people are using it.

I'll give you a good example. Boomy's being used in tons of classrooms, which was not necessarily intentional on our part, it's just we saw all these teachers and we're like, "Hey, what are you doing?" And what we learned, I think, again, you're coming at it from this point, and I hear this perspective a lot, that is fundamentally, "This is a machine producing music." And that's the thing that I just don't see.

Richard Kramer:

Right.

Alex Mitchell:

Because if you're in a classroom and you're teaching music production or you're teaching music composition, one of the things that we've heard is that it's really hard to get students to basically get all the way through. There's so many different things that they have to learn, even if you show them a digital audio workstation. And so when you start a class with Boomy, with saying, "Hey, here's a simpler interface, sliders, BPMs, instrumentations, and you can start from a fully completed song." And invariably what happens is they go, "Whoa, I want to do more." "Okay, let's graduate to a digital audio workstation. Let's graduate to learning more about music." I very much look at what we're doing as an on-ramp.

Richard Kramer:

Yeah.

Alex Mitchell:

Not only to express yourself with music, but also to maybe participate in the music industry in a greater way and care in a greater way. And if there's a model that's been trained with Bach that helps you get there, I guess it's-

Richard Kramer:

It gives them education.

Alex Mitchell:

Well, sure, but also that education can happen in a classroom, it can happen on your phone.

Richard Kramer:

In the bedroom.

Alex Mitchell:

In the subway. It's not as if we are out there to create a better Bach.

Richard Kramer:

Okay.

Alex Mitchell:

It's that increasing the participation that I think necessitates the... Would you rather we not have Bach?

Richard Kramer:

I think your point about increasing participation is great. I have one more question and I'm going to throw it back to Will, but what I fear, and broadly with all this AI, is it comes down to access to compute resource and the immense piles of money and resource pools that are owned by big tech.

What stops these big tech companies from listening to all those 16 million tracks and training off of the songs that Boomy throws onto any one of the streaming platforms to write better, cheaper, faster songs or come up with more efficient tools? How do you think about that always evolving towards monopolization digital economy that we have?

Will Page:

Tech eating tech.

Richard Kramer:

We have had so many promises of democratization of the creator economy and it all ends up back to the peak of the pyramid and nothing much for the long tail.

Alex Mitchell:

It's a really good question. Let me give you a dinner party answer.

Will Page:

You're serving dessert by this point, you're serving dessert.

Alex Mitchell:

Which is a story, and I've got to protect the identities here, but I was in a conversation with somebody at a DSP and it was a pretty intense conversation about partially that subject. And I guess my base answer to your question is what prevents big tech from all that stuff that you said, hopefully for us would be regulation and copyright and the fact that we're pretty unique because we're a music company that is also developing AI technology that has music rights, but also is using our own rights and wants to license other rights. We're in a lot of the middle of this stuff. It's pretty fascinating stuff, but I think there's a lot of sort of pontification, there's a lot of talk, there's a lot of panels.

I was talking to this person and we shared a slight kinship over... We have to make those decisions. They have to figure that out. And something that was really interesting that came out that conversation was this notion that... Let's look at the history. You've got... Let's just pull out 2 Live Crew as that very famous case. And you had a situation where, I'm not saying it's exactly the same, I'm not equivocating, but just in terms of the history, you have a situation where there was an artist, they created some stuff. It was considered very offensive, it was sampling things without permission or whatever, and it was banned. It was actually banned.

The legislature got together and they banned this album and they pulled it from stores. And there was a decision. And imagine a universe where that was it, and then basically hip-hop probably, and a whole gigantic cultural movement, would essentially maybe not have ended, but definitely have gone a different way if it had just stopped there with, "Okay, that's too offensive, that's using uncleared samples. Can't sell it, you're done."

Of course, that's not what happened. Of course, there was a First Amendment case, a very famous First Amendment case that overturned it and said, "Look, at least in this country, we believe that people should be able to express themselves in music, and you have to be able to sell this." And it opened that door.

I think what's interesting, and I'm only making this point about big tech because you brought it up, but I don't think that's the way that things are going to go in this way. I think it's going to be people inside of tech companies and inside of DSPs now deciding what is and what isn't going to go forward.

When I look at the role of big tech, I don't look at it cynically. I look at it as the reality is the consumers have oriented around a relatively small number of platforms, the policies of which are going to impact the market in a pretty outsized way.

Will Page:

Good comment.

Alex Mitchell:

I don't think they're sitting around over piles of gold saying, "Oh, boy, how can we be evil?" I really think they're, from personal experience, they're struggling with some of the same questions that you are, that we we are, about what this future is going to look like. I think there's far more questions than answers at this moment in time.

Richard Kramer:

Absolutely.

Alex Mitchell:

But these perspectives are really useful there. I do think it's important that big tech understands its responsibility in being sort of the... It's going to be their decisions, what to let on and what not to let on, and obviously AI is a big topic in that realm right now.

Will Page:

And it's interesting when you say struggling with these questions, just like perhaps 18 months, two years ago, they were struggling with questions on content moderation. And I always like to quote the famous line where Mark Zuckerberg announced to Wall Street that he had 10,000 people working on content moderation and the stock price fell. That's a great symbol of America. "What? You're taking this stuff seriously? Dump the stock." I think he now employs more than 60,000 people in content moderation.

We're going to move to smoke signals. Rich will tee you up for that. But to keep our listeners listening as they go to desserts and after dinner coffees, fun fact, you talked about not being an expert in what content is now in the public domain. And Richard, your point about Bach is really important, and we're going to talk about Benjamin Britten, which is still in copyright, that's one thing. And we're going to talk about Bach and Beethoven, that's a whole different thing.

Did you know which composer and which compositions, well famous, went into the public domain this month? Tchaikovsky's Romeo and Juliet. Go get Boomy wrapping that up on Monday morning.

Richard, time to get on our smoking.

Richard Kramer:

Yeah. Trying to throw our guests the question of what are the smoke signals, the kind of things that you hear that make you go, "Uh-huh," in this debate about whatever topic we're having our podcast on, in this case, AI and music.

What are the one or two things that people say about AI and music, and maybe I've inadvertently echoed them, that are just the most irritating inanities that you've come across and just say, "Well, no, you just really don't get what this is all about"?

Alex Mitchell:

Believe me, I could probably rant for an hour on that.

Richard Kramer:

We're not going to do that.

Alex Mitchell:

But I'm going to start with-

Will Page:

We'll have lung cancer issues here, so just take two puffs on that cigarette.

Alex Mitchell:

One of them, and I really don't mean this to push back on you at all, but you said it, so I want to bring it up, is legitimate artists. That's one where I start to wince. In part because it's such a complicated question. And again, I clearly have a bias. I have competing biases. I have a musician bias, I have a policy advocate bias, I have a Boomy bias into that question, but I think what is a legitimate artist that, quote, unquote, belongs on the shelf, and saying that, "Well, because that's different or it's probably not going to chart," the idea that someone's going to decide that because of the way... And this isn't existential by the way, this is real time, that, for example, because of the way a musician creates their song, they should be paid differently, is a totally unprecedented idea. It would be akin to saying, "Well, look, you've got auto-tune in that song. That's not a legitimate singer. Pay him 25% less or maybe don't even include them along with all these other real artists."

Richard Kramer:

Or a harp is more expensive to record.

Alex Mitchell:

Right. "Well, you used a synthesizer for your orchestra. You didn't hire a real orchestra. You should probably pay that differently."

Richard Kramer:

I guess on this question of legitimate artists though, I was trying to draw a distinction between companies that are simply generating music for commercial intent with no real artistic aims versus the thousands and tens of thousands of people using DistroKid in their bedrooms to upload music because they're passionate about it. And the illegitimacy is really in the subversion of the commercial model of music, to extract money and for none of the passion that you expressed might be between parents and kids writing their own bedtime song.

Alex Mitchell:

Yeah, sure, but I would also say that I've heard some music from those legitimate artists, some of which seem... I want to stay neutral and I don't want to call anybody out, but I had a moment, and I have to stay vague here, where I was listening to a song by a rapper who I guess one could describe as a mumble rapper.

Will Page:

That narrows it down then.

Alex Mitchell:

And the lyric was something to the effect of... They referenced, and I'm trying to keep this vague, because I'm, again, politically neutral on this stuff, but they made a mistake in the lyrics, let's just call it a mistake. And the mistake was so glaringly obvious and created such a... Obviously this person does not have an understanding of instruments if they're going to make that mistake in lyrics. It's also kind of incredible that it got all the way through the release with that mistake in the lyrics. And that had a huge amount of streaming volume on it. It was on the charts.

As a violinist. I think frets are cheating. I think you can draw that purity line anywhere you want. I think if you drew that purity line through the people who are making music for commercial intent, I know a lot of artists too, yes, they're expressing an artistry, but they want to make money off of it, and some people see volume as a way of doing that. I'm talking about artists who are signed to labels who you would totally put in the legitimate category.

I just don't think it's that simple. And that's why it's a smoke signal to me.

Richard Kramer:

Right, okay.

Alex Mitchell:

Is that I think it's a door that opens towards this is real, that's not real, this counts, that doesn't count. And maybe that's where it lands, but we should at least understand that's a totally unprecedented idea, not something that anybody's ever talked about before.

Richard Kramer:

And by the way, right now, there is a complete parallel, which we should have another podcast on, in the news industry because there are dozens of AI generated fake news websites, which are just borrowing content.

Alex Mitchell:

Oh, sure.

Richard Kramer:

From other websites to simply attract clicks and revenue. And that's what I see is illegitimate, if you will, another smoke signal.

Alex Mitchell:

I would say, you hit on them in the very beginning of this podcast, dilution bit, I think too, which again, similar, but this idea that you're diluting the pool or that it's the fault of a content creator or it's the fault of a kid in their bedroom, to borrow your phrasing, that there is a lot of content coming out a day.

I've been having a lot of conversations recently about the, quote, unquote, 120,000 songs a day problem. It's like, "Well, what are we going to do about the 120,000 songs a day problem?" And this gets me in trouble sometimes, but what's the problem? Why is that a problem?

Will Page:

True.

Alex Mitchell:

It's also a small number. Remember the photos number is hundreds of millions? And this is something that more and more people are going to be doing. And to get to something that you said, actually, I think more in part one, I don't think this is going to really impact what we would call, or maybe what you would call traditional or pre AI or whatever, legitimate, to use your terms, I don't think this really affects that market. I think this grows the entire market.

This fear of dilution is really just based on how I think the economic reality sits today, but that's going to change. It always changed and it'll continue to change. And I do think this will be a forcing function for some of it, but we have to be thoughtful about how we accomplish that. That's another one that just kind of makes me, "Dilution? It's indie music." I'm never going to see indie music as being dilution. And personally as somebody who came from that world, it's not for me to judge, I guess.

Richard Kramer:

Yeah.

Alex Mitchell:

What counts and what doesn't count.

Richard Kramer:

And there's an incredible and growing role for, and there was a brilliant interview with him just recently on Music Business Worldwide, DJs like Gilles Peterson who have 150 or 200 or 500,000 records to choose from, and they have three hours curate [inaudible 00:44:26] on a Saturday afternoon.

Alex Mitchell:

Sure.

Richard Kramer:

And you want to listen to those three hours because you know that selection, among all of the stuff that Boomy and everybody else puts out, is so valuable.

Will Page:

Yeah.

Alex Mitchell:

Yeah. To my knowledge, AI has not impacted that culture.

Richard Kramer:

At all, yeah.

Alex Mitchell:

In any way.

Will Page:

Alex, if I can take a minute to bring it all to a close, and the first thing I want to say is [foreign language 00:44:53], thank you, on behalf of myself, Richard, and our audience. And our audience isn't Harry and Meghan Big, nor is the money that we're making Harry and Meghan big, which is, by the way, way more money than was reported, that's for another podcast. But I would stress, usually this podcast is Richard Kramer taking to task bankers on Wall Street, but we've had Jessica Powell from Audioshake.

Alex Mitchell:

Love Jessica.

Will Page:

Ed Newton-Rex, the founder of Jukedeck.

Alex Mitchell:

Yeah, yeah

Will Page:

And now yourself. We've possibly got the podium of gold, silver, bronze, AI musos to come on the show and set out their stall.

Now, when you set out your stall, I just want to wrap up with two comments. One, your point about education is huge for me. If you're going to hire a VP of Policy, I'm sending my resume to Boomy tomorrow because there's one way that you could prove the effect of Boomy, which is do classes which involve Boomy see children maintain a music education for longer than it would otherwise be the case, that is, at least in the Scottish education system, once you get to 14/15, you can drop music and never study it again. Would a classroom that has Boomy see a lot of those kids who would've dropped, pursued to 16/17? Would be an amazing policy angle for you guys to explore.

Second and closing point is one I've reiterated on this topic before. Capitalism was described to me by Alan Blinder, famous Federal Reserve Governor when I was at university. He said to me, "Capitalism is when you employ a gardener to cut your grass because you can do something more productive with your time." Look at the GitHub Copilot case. It's a threat to coders. Now, coders see it as an opportunity. I can code four times faster thanks to GitHub Copilot. I wonder whether AI, in the broadest term, is just one big F-off gardener, and what Boomy is doing is providing that gardener for the form of music creation into the future?

Alex, huge thank you for coming on Bubble Trouble, man. You've set out your stall very impressively, and I want to get you back on because we've got so many rabbit holes to explore here. But thanks for giving our audience an introduction to what Boomy's doing, and good luck with the next 16 million songs that's going to come from the platform.

If you're new to Bubble Trouble, we hope you'll follow the show wherever you listen to podcasts. And please share it on your socials. Bubble Trouble is produced by Eric Nuzum, Jesse Baker, and Julia Natt at Magnificent Noise. You can learn more at bubbletroublepodcast.com.

Until next time, from my co-host Richard Kramer, I'm Will Page.