Switched on Pop - Can AI 'Algorhythms' Write Pop Songs? (With Taryn Southern)

Episode Date: May 3, 2018

There is a lot of scare about the impending future of artificial intelligence making humans irrelevant. Musician Taryn Southern examines this narrative through her song “Life Support,” written with the aid of AI composition tools. We dispel current myths about AI music and discuss its future opportunities. In the second half of the show we run a musical 'Turing test' to see if you can identify music made by a bot from that composed by the hand of a human.  Learn more about your ad choices. Visit podcastchoices.com/adchoices

If you're tired of endless scrolling to figure out where to eat, same. I'm Stephanie Wu, editor-in-chief of Eater. We've just launched the new-ish and way better Eater app. It has all the restaurants we love, gives you personalized picks wherever you are, and serves up smarter search results just for you. You can find my list of the best places for martinis and fries in New York City. And save your favorite spots, share lists, follow editors, and book right in the app. Download the Eater app at Eaterapp.com.

It's free for iOS users. Welcome to Switched on Pop. I'm songwriter Charlie Harding. And I'm musicologist Nate Sloan. And we are very excited to have joining with us today. Terran Southern technologist, actor, songwriter, extraordinarily talented person who's here with some new music that we're going to deconstruct today. Welcome to the show. Thank you. Thanks for having me.

I just want to get right into this because I think that your music is going to overwhelmingly surprise people. It's perhaps even going to be. going to frighten some listeners. And I don't want to ruin this surprise. I figure what we could do is just jump right into it and hear a little bit of your new single life support. Awesome. Yeah. Cool. So, turn, this is a really fun song. Thank you. That's me. What do you think is most going to surprise people about this track? I think that the most surprising aspect of this track is that it was composed using only artificial intelligence. Composed and produced, shall I say. Absolutely. extraordinary. Wow. A robot made the music. Yeah, well, I mean, that's like the quick and easy, I suppose,

marketing definition, but I'm sure on this type of podcast, we're going to get really into the dirty details of what that means and what actually was done using the AI. Totally. So your upcoming album, I Am AI, is composed with many of the latest artificial intelligence music technologies. And today, what I want to do is break down how those tools assist the creative process. and dispel some of the myths, which I might have set up in a completely inappropriate way at the top of the show, about how AI may be replacing humans or not replacing humans, specifically through the musical process. I'm glad that that's how you set it up, because that's exactly the best way to unpack all of it, is to start with the headline and then work your way backwards.

Right, exactly. So we got a taste of the chorus already. I want to start from the top of the song and listen to the verse. Distance, what have we become? Fated lights, desert skies, have we come undone? Consumed by a game, designed by the blind, there's no way to win, no escape route I can find resign. This is really powerful for me because in researching the work that you've done, you're a bit of

an AI enthusiast. You consult and support people and learning more about. about artificial intelligence and its role in modern society. And here, your music is taking sort of a more dark dystopian point of view. And I wanted to know, why did you choose to take on this narrative point of view as your point of inspiration? Well, the album is entitled, I Am AI. And so the entire journey that I take through the album is one of exploring, I think,

all of the opportunities and perils that are presented by technology. So this song definitely takes a more dystopian view. Although ironically, I wrote this song one day while feeling particularly down about the obsession that people, particularly young people have, with social media. So I was actually more projecting our current state with technology that I'm actually hoping that artificial intelligence and other technologies will assist us in pulling ourselves out of. That's beautiful. By demonstrating that it's terrible for us. I think we may share some similar feelings on that. What are some of the issues that stood out for you on that day? I think that we've become obsessed with likes and followers. And this is not just amongst social media celebrities. This is every young teenager in America who now perceives their own value based on how many people are looking at and liking their posts. And I think that there's just such a big danger in growing up in a time where you don't see the world.

through the eyes of how it is. You see the world through the lens of Instagram and how others view you. And I think that it's so dystopian to me. I feel so disheartened by it. That is terrible. Yeah. And as someone who grew up in YouTube culture, I mean, I was one of the first people on YouTube and on Twitter and on Instagram.

And I'll be the first to say that I think it's probably one of the worst things we can be doing to ourselves. Wow. That really is a terrible. situation in which we do find ourselves today in these filter bubbles, which you actually reference in your music. But it's also a wonderful point of creative inspiration. You, in life support, your single, use these issues as sort of double meanings. On one hand, I'm hearing this narrator, which almost has a matrix-like experience of being tapped into a game and trying to get off

life support. But it also has immediate connotations to what's happening today. in our addictions. In your verse, you open up, you say, consumed by the game, designed by the blind, no escape route, I can find, I'm resigned. And I think this is really brilliant because there's a third meeting here, which is that you're actually referencing how you designed this music, designed by the blind, without any sort of creator behind it. Yep, good job. You're the first person to point that one out. Oh, thank you. Charlie just got his comment liked. You just got a like and a follower, Charlie. I feel validated. I really hope that gets re-treated a million times. There's some really beautiful lyrical choices you're making here that are present, past,

future, referencing all of the challenges with technology. What I really want to do today is talk about your sound palette, because I think that's where things get particularly interesting. And you've chosen, obviously, a very deliberate sound palette to evoke this dystopian sense. And I was hoping that you could describe what is the soundscape that you want to to establish to create the scene? Well, to some degree, it was a bit dependent on the tool I was using, which in this case was an AI software called Amper. And I found myself really enjoying the cinematic soundscapes that Amper was creating.

A couple of the guys who started Amper actually came from Hans Zimmer's team. And you could tell because the soundtrack sounds and library were just, I think, so full and rich that I found myself perusing that side of the software far more than some of the pop. and rock stuff. And so inevitably, I was conjuring up various images as I would listen to the music that was being thrown out at me. And there was some variation of this piece that was the first variation of the piece that felt dystopian and had elements of hard clanking and metal sounds in it, which kind of felt technology. And so I started writing this song based on that and then probably iterated on it, I don't know, 20 to 30 times. And then once I was happy with a melody,

then I would start bringing another instrumentation, as you said, with the intention of creating a soundscape that is dark and electronic and has a mixture of old school symphonic with just very hard, percussive sounds. Okay, I feel like we have to back up for a second because we're talking about ideas of composition and intentionality. And yet, hold on a second, you're using this tool called amp or music to assist you in the composition process. So could you take us back and just tell us what is Amper music and what is it doing for you as a musician? Sure. So Amper is one of many tools that are currently on the market that uses artificial technology in some former fashion, machine learning to create music. And it's a tool in that any kind of artists can use it and iterate on the music that it provides. And in doing so, the machine learns. and supposedly, or is supposed to, get better as it goes on.

And as the artist, you essentially set parameters and iterate as many times as you'd like in order to get a result that you are happy with. So I actually grabbed a few clips from Amper. And I thought maybe we could play some examples of things that you could choose from Amper to ignite inspiration as a starting off point for a song. So the first track I've got here is a selection of a cool hip-hop. Hotbeat, light and spacey, 70 BPM in the key of A. I've decided, well, perhaps I don't like the key of A.

I want to go to the key of C. I want to speed the whole thing up. I select all those little settings and boom outputs the next version of the song. And I don't know about you, but already I am liking this track a whole lot more. Yeah, much better. It's got a little sauce to it. It does have a little sauce to it. I'm not in my head.

and if we decided that we didn't like hip hop and I wanted to go into maybe say a 90s pop feel with a driving sort of sense speeded up even faster we could get Amper to output this that was the wrong choice my sixth grade self is very excited about that song I think in playing these sounds it gives listeners a sense that

Emperor is not outputting a complete song It's giving you a point to begin with. That's right. It's giving you pieces. And the way I liken the songwriting process with Amper or with actually many of these tools is you become more of an editor or a director rather than a composer.

You're basically given a ton of raw material and it's your job to shape that material to cut that material into something interesting and transform it. But you're still using the source material from the AI composition tool. So after listening to a bunch of these Amper tracks, I would imagine that a lot of people

would be asking what is the artificial intelligence component? And so what I thought we could do is break down some of the things that an artificial intelligence might be doing. And in order to do that, there's this really great open source music AI tool called Magenta,

the Google Bill. And they've got a handful of tools that help explain, I think, how a computer is being used to generate music. So I thought what we could do is just walk through a couple of examples

of how artificial intelligence can create sound, it can create a beat, it can create a melody. Great. Cool. Obviously, Taryn, you're very familiar

with these sounds, but this stuff is totally new to me, and I'm guessing absolutely new to Nate. And so I find it absolutely fascinating. And the first one I saw, I've heard that you've composed with N-Synth is this really cool new form of synthesizer, where two totally disparate sounds

can be combined to create one new entirely unorthodox sound that's never been heard before. So I have a clip for us of, well, I think you'll know the instruments when you hear them. That's a cow. That's an organ. And that's an organ cow. A corgan. A corgan. That's a beautiful thing.

I never knew I needed one. Wow, I did not see that coming. Okay. Fascinating. Yeah, I used it to combine the sound of my cat with a synthesizer. And what happened? It's a ridiculous sound. And of course, the only person who cares about that sound is me because of the story behind it.

Well, and your cat. Yeah, it doesn't always create a beautiful sound, but it's all about the narrative, right? Absolutely. And also, these tools, they're there for the happy accidents. They're there for those inspirational moments. So I love that you've got a cat. I don't know what a cat synthesizer should be called a cat synth. It's a beautiful thing. I bet it made a great lead line in one of your songs. Yes, exactly. So the other thing that

surprised me about Magenta was figuring out how an artificial intelligence may help make a beat. And there's this great tool they have called Beat Blender. I think this is a really great example of how software can, in a fairly obvious way, take two different things and again to turn it into sort of a composite. So on B-Blender, what you can do is you can basically insert two very simple beats and see what they sound like if they were matched together in a random form generated by a computer. Simple drumbeat, number one. Even simpler kind of just straight up rock beat, number two. Yeah, pretty by the numbers. And you put it together. And if we had asked Google's Magenta to create five more variations of the same thing, it would try to do so.

And I think here we're seeing both the magic of artificial intelligence trying to figure out what notes to play when and where and also some of the limitations. I don't know. What do you all think about those beats? They probably wouldn't be playing the third beat at my wedding, but it's a good example of how an AI is attempting to make sense of two very disparate pieces of music. Yeah, like Taryn, I don't know that I would ever play that again and play comedy. But I could see that as the basis for a new composition. And what I did find exciting about it was that it was not something I'd heard before. Oh, right. That was kind of cool.

It was familiar, but nothing I'd heard before. And that's rare. So that I did think was cool. Okay. So if we've got sounds, we've got rhythms, I think it would be only appropriate to look at melodies. And Magenta gives us one more interesting tool to look at it. An AI, which will take one melody and like the beat tool. take a second melody, and find some sort of in-between of what you might infer if you like both of those.

So let's take a listen to a twinkle-twinkel little star combined with an arpeggiated sort of synth line. That's really unpleasant. Okay, so those are the two different sounds. And then if you play the next clip, you'll hear what happens when you try to get something somewhere in the middle. I was feeling that. I like somewhere in the middle. It had a little something going on there. Yeah.

It's interesting because the question I always like to ask myself if I'm able to talk to the engineers or the people who create the software is like, what does it mean by having the machine take something in the middle? And how do you program that? And why is that perhaps the most valuable way to get something interesting? Like what are the techniques that humans use to pattern match good music? And how can you potentially apply that type of human learning to AI so that you actually. actually get good results, you know, rather than just total randomness. This is where having listened to some of these magenta tools, I'm like, okay, I kind of get what something like Amper is approaching. It's taking the sound components, the melody components, and the beat generation components of AI and sort of putting them all together to then create some sort of composition. What I'm hearing underneath what you just shared, though, is that there are some limitations to what's possible. So could you frame where AI, you know, right now is particularly useful in making a composition. Absolutely. I think it's particularly

useful in providing interesting inspiration points. You know, if we look at the way that human composers write music, it's fair to say that we often compose in the style that we are accustomed to listening to. And it's rare that we go outside of a given box. Right. And so the opportunity that I think AI affords is offering up styles and possibilities that, like, one of you might have mentioned when we were listening to the clips, was a bit surprising or unlikely. And then taking from that and building upon it. And I think that the opportunity that we all have as musicians, composers, and whatnot, is to see these kinds of tools is actually an inspiration point that maybe exceeds our own abilities or our own ways of thinking, our own limited ways of thinking.

that is where I see the biggest opportunity in it currently, because the reality is most of these AI tools are not making great pieces of music on their own. They require a fair amount of retooling or editing, so to speak, which probably makes a lot of human composers and musicians rejoice. But the reality is also that that may not be the case in five to ten years. It may be able to fully make music that is quote unquote good on its own. But at the moment, that's a subjective analysis. And so until we can force ourselves out of our own boxes and start creating, I think, music that surprises and astounds and maybe is just totally different than what we've heard before, AI will only just learn off of our own creations. Well, that gets me to something that I'm

really excited to get to share, which is, even though there are limitations, with some human intervention, AI is making some pretty spectacular music right now and I thought it'll be fun to see if our own ears could outsmart a robot and that's exactly what we're going to do when we come back after the break. Convierte your passion in a business with Shopify and bathe records of ventas with the form of pay with a better conversion of the world. Has you heard it? The mejor conversion of the world. The incredible system of Pago of Shopify

facilitates on your site web, in the networks, and in whatever place. That is music for your ears. No, let's more vultas. Your negotiation your negotiation will be a super-exit with Shopify.

Empeza to period of a month in Shopify. orgs. Immigration may be Donald Trump's signature issue. President Trump is now

targeting predominantly Democratic cities for ice raids and deportations. Dozens of protesters clashing with immigration

and customs enforcement agents in Minneapolis Tuesday. We will begin the process of returning millions

and millions of criminal aliens back to the places from which they came. But what we want to do in this space is talk about America and politics beyond the current president. So what do most Americans think about deportation and border security, period? I think that Americans are definitely against the kind of violent displays that we've seen in the street from ICE. When it comes to the question of deportation, the answer is more complicated. My sense is that people want border at the border. They don't like the idea of having no idea who's coming into the United States at any given time.

The view on immigration from the bottom up instead of the top down. That's this week on America Actually. Every Saturday in your audio and video feeds. I have designed a really fun game, a sort of non-scientific touring test, if you will, to see if AI music can outsmart a PhD and a brilliant musician. So Nate, Taron, are you ready to see if you. you can outsmart AI music. I'm ready and terrified.

Okay. We've got five clips. Those listening along. See if you can keep track. See if you can win. First track. We're going to start in the classical sphere. We're going to listen to some Bach.

There was a wonderful Bach tool called Bach Bot. And Bach Bot asks you to listen to two pieces of Bach. One is Bach and one is not. And you've got to figure out which one's the bot. Okay. So that's the first half. Let's try the second clip and see which one is human, which one is bot. How are you all feeling?

A little nervous just because of what's at stake here. Wow. This is pretty good. So, Tara, let's start with you. What do you think? I would say that the second clip was created by a bot. Nate, what do you think? I'm also going to guess that the second clip was created by a bot.

Second clip was the Bach bot. Wow, but it's so close. But it's subtle. It's very subtle. It's really subtle. In some ways, I actually thought that the bot one was more interesting, but there was something about it that felt slightly unbock-like. And I don't know what that, you guys would probably have a better sense of what that would be. Well, so Bachbot is trained on a neural network, which basically inserts all of box compositions, take all the music, which is no longer under copy, right? You can put all the lead sheets into a computer, scan it, and that computer is going to figure out which notes should most likely come after. a note that came before it and do that across all four voices and brilliantly it prints out things which sound a whole lot like Bach. However, not quite Bach. I don't know about you all,

but I'm curious. How did you decide that the second one was the bot and not the Bach? There was almost a little too much variation that then struck me as a statistical generation of notes as you were saying versus a cohesive arrangement where there's potentially more repetition of certain notes to create that cohesive performance. That was my sense of it, but again, I thought it was still very, very close. No, I had a totally similar thought process. The repetition in the Bach feels very familiar. And I would say not even just the repetition,

but then where Bach chooses to do the variation, feels very human somehow. Not random. And I'd have to think about why that is a little more, but I'm definitely quaking in my boots a little right now. Things are about to get harder. We're going to move on to more symphonic sounds. That sounds like music.

That sounds like music. Yeah, right? It's good music. The question, though, was it composed by a human or a bot? Let's listen to the next clip and compare. Do you feel like you have enough information to make a conclusion? This one is hard. This is really hard.

Was this with Ava? This is with Ava. Tell us about Ava and why Ava might be confusing us. Well, this one's going to be tougher to figure out, because Ava uses, and I'm working with Ava. Actually, my next song, next week that is being released is with Ava. Ava uses human instrumentalists to transpose the music that's composed with the AI. And so you're going to hear, there's a human touch to it.

It's not like it's amp or just spitting out the finished production, so to speak. Yes. These are both real orchestras, real humans playing both of these tracks. And so what we're trying to hear underneath it is, is the underlying composition, human war machine. Oh, I'm going to mess this one up. This one's a coin toss. You may notice that I intentionally chose two songs,

which use almost the exact same melodic motif. It might hurt you to know that you've heard one of these pieces before. What's the composer of one of the pieces? Can you say? Oh, I don't want to give it away. You have to decide first, and then I'll let you know. Okay. I'm going to guess that the first piece was the AI.

Nate? I'm also going to guess. The first piece was the AI. The first piece was Ava, and the second piece was John Williams playing Ray's theme, which is the theme of Ray from the new Star Wars trilogy.

How did you distinguish the difference? Or are you just guessing? I felt that there was more build to the second song that started off very sparse and then slowly got bigger, whereas the Ava track had a lot more going on.

And also just from working with Ava, I guess I've felt that there was, I don't know, something about it, something about the simplicity of the second song almost made me think that it had to be the human, because Ava, as a result of having a lot of data, it tends to inject a little bit more in the pieces than you might find with a human composer, even when it's learned off of a composer that's relatively simple. Fascinating. I'm inferring that you're saying that Ava has a personality and a musical style, which is terrifying. Nate, what did you hear?

Yeah, well, I wouldn't have put it in the same terms, but what I did hear in the first example was what sounded like almost sort of a random doubling of the initial line with different instruments first, like some kind of high Glockenspiel and then maybe some low strings. So it's being doubled by all these different instruments, which might not be unusual in itself, but I think the way that it went from high to low just felt kind of arbitrary rather than it was building towards some inexorable conclusion. which as Taryn pointed out is more happening in the second one. It has this very clear goal that is trying to reach toward. But man, I was definitely sweating on that one. Okay, so we're going to take you from the classical realm into the more modern realm. Let's take a listen to the next clip. And again, we're listening for Bot or Not.

That's the first clip. There's the second clip. So what do we think? Which one's the human? Which one's the AI? This one's not a fair fight because I actually know the answer because I know the AI track. So I will stay out of it. Nate.

I'm going to say that the second one is real. Sorry, whoa. They're both real, aren't they? The second one is, yeah, sorry. The second one is made by humans. Okay. And the first one is made by AI. I hope you're not trying to pull my leg here because this is deeply embarrassing.

The first one was actually a composition by. the Beatles and Giles Martin. I was going to say it sounded like Tomorrow Never Knows Meets because. But I thought, here comes the sun backwards. This sucks. I don't want to play anymore.

Nate has read basically every biography of the Beatles. That was a basically remixed track of the Beatles that George Martin's son, Giles Martin did for the Vegas Cirque de Soleil show, Love, which makes us wonder, what was the second track? Terran, can you tell us a bit about it?

Yes, the second track is a track off of the debut album of Frank Wah Benoit who built Flow Machines, which is an AI composition tool that Sony acquired and then later Spotify. That is correct. This is a song called Daddy's Car. It was composed using styles of the Beatles.

Yes. And in this case, it was actually, again, sheet music that was written and then humans playing the actual music. And this is basically a interpretation of what an AI would think the Beatles music

would sound like. And so, Nate, you are fooled, you are shamed, you have lost your PhD. And you can... Taryn will be resuming my duties as co-host for the rest of our podcast.

So Taryn is currently three and O and Nate is two and one. So the stakes have risen. We're going to move on to the next track and this might get even more difficult because we're going to move into the realm of electronic dance music.

Let's figure out human or AI. And we're going to pause there and we'll go to the next clip. That's what you got. You've chosen your clips very wisely. I also have to sit out this one because I know the AI music. Oh, interesting, but you don't know the human music. No, I don't. Oh, okay.

Nate, this is your... I know neither the human nor the AI. So Taryn can't win a point here, I don't think. So I think we're going to have to see, what do you think? I think the second one was human and the first was AI. And what's leading you to that conclusion? I don't know. I really have no idea.

I just like the second one a little more, so I'm hoping it's a human. Yeah. The second one is a track by Moby off of his new album. Oh, funny. Yes. The first one is Taryn. It's an Amper track that they have a template for. They have like 25 templates on their site that you can use to start a track and I recognize this template.

All right, moving on to our last song in the challenge. Whoa, that's cool. And let's listen to the next clip. Again, we have to figure out which one is the AI, which one is the human. Any guesses? I have to sit this one out because I know the human behind. You know their music. Well, this is great. It makes sense. You are an expert on artificial intelligence. So, Nate, you're alone again, and everyone is looking at you. All lights, bright. Can you redeem yourself? No pressure whatsoever.

This is so embarrassing. Not only has this game exposed how much more musical knowledge, Jaron, has than I do. It's exposed how bad I am at guessing. I'm going to go with the first one in this case, because while it does sound kind of random and insane, it also has these moments that seem so perfect. perfectly calculated that they could only be made by a human brain, whereas the other one seems like it could have gotten lucky. Well, this was kind of a trick question. Taryn, you know the artist. Who is this? Andrew Huang. He's a YouTuber slash musician.

And he's extraordinary. And in this case, he's actually composed both pieces. However, the first piece was composed entirely of sounds from En synth. that he put together with the team at Google. And so these sounds are sounds that basically nobody has heard or created ever in human history. And I think he took something like 200 different clips, mashed them all together, made these weird, strange synthesis mashup sounds and then made a composition from it. And he made one that sounded both beautifully electronic and disturbing and dystopian, but also very human. Wow. Okay.

So cool. Andrew's super talented. So talented. His YouTube videos are definitely a guilty pleasure of mine. Okay. I'm not sure who won. This is difficult. I think we know that Terran clearly knows her AI and Nate doesn't know what he's talking about whatsoever. So we're going to hand it to Terran. Congratulations. You have won the AI or not a challenge. Thank you. I'd never win anything. So this is very exciting. Well, I think that this leads us into a really important conversation about what are the limitations of this technology? What are the ways in which we are identifying what is made from an artificial intelligence and what is made from a human? What are the things that you've uncovered in your creative process, Terran? Well, I think to some degree, as you heard in the clips, and not all of the clips were solely AI in the sense that as we discussed, some of them were produced using human musicians. and in many cases, such as the one with flow machines, the Beatles song,

there may have even been a fair amount of even notation editing happening in the back end. Absolutely. And that's not a bad thing. I'm just telling you that that's what's happening. You're saying that my sample set is a little bit misleading, and that's true. I wanted to play. No, no, no, no. It's more that the marketing of what's actually happening behind the scenes is misleading.

I'm not criticizing any artist. A lot of it is just even press and the press that I've dealt with in talking about my album. It's easier for people to slap a quick headline onto something rather than actually explain the process. For me, I think that just the process of using AI as an inspiration point

is super interesting, but it's probably easier for a newspaper or a blog to say, like, song made with AI, you know, and then they call it a day. As we heard, it's tough for AI to, I think, build a song that has built, actually. Right.

Any sort of sense of development, melodic development, harmonic development, and consistency at the same time. It's like in some ways, doesn't it feel like it kind of reaches too far? There's no patterns. And yet humans would be, in some ways, as you put it, simpler, which is almost unexpected. Yeah, the AI needs to learn to calm down because it just gets a little too excited and it throws a lot at you. and sometimes it's like being punched in the face.

And I think that's fairly consistent across pretty much every platform I've worked with is it just tends to be pretty busy. And there's just not a lot of coherence to a particular song. It's hard to stay in a specific mood or for it to exercise any sort of self-restraint. That's the limitations, really. There's clearly so many things that humans are better at. Currently, we're better at pattern matching in music. We are better at reflectivity and reflecting back on the music and asking questions about it.

that's a real big general AI problem that I don't think is even close to being solved by musical AIs. I think we're also better at making happy accidents, but putting them in intentionally. There's all these ways that, especially when I'm listening to that Bach, which I think is obviously the hardest to tell the difference, both perhaps because of our familiarity with the music, except Nate, you mean, you should really know what you're talking about, the whole PhD thing. You're getting burned today. I'm really sorry. But aside from the Bach, I mean, usually what I'm looking for is any sort of sense of development and form. And I'm just not hearing development and form. But to your point about marketing, sort of jumps over the point, which is that these AIs are built for a very specific purpose, which is right now that they're built to make commercial music for content, right? They're basically replacing the most boring Muzac in the world. I love that. You should actually tell the AI companies to make that their new marketing headline. Replacing the most boring music.

in the world. One stock music at a time. Things that are just like, I've made an Instagram video and I need some sound in the background, which sounds vaguely like music. AI can do that fairly decently at this point. Yeah. I made kind of a bad joke recently in an interview and I felt sort of terrible about it. But since you said your funny comment, I will share it with your audience. When someone asked me, they said, are you worried about AI taking composers jobs? And I was like, as of right now, only really bad composers. And then I thought, oh man, I hope they don't quote me on that because that just might look bad.

But if you just take it as it is by itself without any sort of human interaction, at best, it's replacing, you know, stock music behind mediocre YouTube videos. Am I allowed to say that? Is that okay? A hundred percent that's okay. Yeah, absolutely. I mean, that's intent, right? This is for people who are making content.

that can't afford to license high-quality music right now. Licensed music for advertisement is not going away right now. I wanted to give Nate an opportunity to redeem himself, if you don't mind, because I feel like this has been a really embarrassing episode for him. I appreciate that.

Nate and I had a really fun adventure recently visiting New Orleans, where we got to see that this issue of humans being replaced by technology and music is by no means new. The marketing hype should not be that exciting. It's an old story.

So, Nate, you have a story for us about this, don't you? Yes, Charlie and I went to visit MS. Rao Antiques, which is this incredible store in the French corridor of New Orleans. When you enter MS. Rao, one of the first things you see is an original Enigma machine. And then you go into the back room, and that's when it starts to get crazy. They have a giant wooden box taller than a person, probably like eight feet tall by maybe six feet wide. And you think, what is this box going to do? And then you flick a switch and it opens up and reveals an accordion of piano drums.

This is what you hear. That's crazy. This incredible creation called an orchestrion. This is an orchestra in a box. This is when your musicians are too drunk to play or they're out of tune or they, you know, hawk their violin at the pawn shop and you still need to entertain people. You turn on the orchestrion and you have a whole orchestra played by pneumatic pumps. So that's the original DJ basically, right?

That's the 19th century equivalent of Diplo playing Coachella. That is all you need. But we can go even further back. I mean, we can go back to, if you're going to indulge me for a moment. We can go back to ancient Greece because perhaps the first sort of mechanical instrument, the first AI, is the Aeolian harp. Are you familiar with the Aeolian harp, anyone? Oh, well, you were in for a treat.

Aolian as in. from aolis, the Greek word for wind, I guess. Also a minor key. The aolean harp is played by the wind. So you just leave it out on, you know, sort of a wind-swept cliff somewhere, tune to a certain, you know, series of sympathetic intervals. And you just wait for Gus to come along. And it creates these fascinating random soundscapes.

Wow. That's so cool. Wow, it's really beautiful. might be better than some of the AI music. The wind does. Make some better choices. The first artificial intelligence. It's not dissimilar.

And then there's probably one more kind of mechanical music that we're all very familiar with, and that would be the music box, which we are often accustomed to in the form of some kind of desktop model that you could hold in your hand and spin. But these were also made on a scale where they could fill up a whole room with their sound. Let's listen to one of these massive music boxes made around at the turn of the century playing a John Philip Sousa track. I remember those. Grandma had one of those. Wow.

Wait, how is this thing work? This is all being triggered by a series of indentations and a metal disc that in turn triggers mallets to strike various size bells. And some of those are very low, which is why you get those very low, and some of them are very high. So we have this huge spread of music. I mean, I know that the technology behind these things is kind of shocking that we could create these, like, robotic orchestras many, many, many years ago. In this form, we have composers like Handel and Hayden and Beethoven are actually

composing for these instruments, for music boxes, for orchestrion. So these are like a real part of the musical tradition. How old were these? The music box we just listened to was from around the turn. of the century, but these were being produced from the 1600s forward. Wow. Whoa. I picked this one that featured a John Phillips Sousa track because John Philip Sousa,

this is the famous March composer who wrote Stars and Stripes Forever. And, you know, in the beginning of the 20th century was like the biggest guy in American music. He was very terrified and furious about what he saw as the coming. computer robot music revolution. Which was? Well, that was the phonograph record. That was the birth of recording.

And in 1906, John Philip Souser writes this essay that you could easily, I think, apply to someone today who was similarly scared about AI composing. In 1906, John Philip Souser wrote, Sweeping across the country with the speed of a transient fashion in slang or Panama Hats. political war cries or popular novels comes now the mechanical device to sing for us a song or play for us a piano in substitute for human skill, intelligence, and soul. There you go. Where did you get that article? That's amazing. This was a long article he wrote in 1906 called The Menace of Mechanical Music. And I'll be happy to send it to you or we can post it online and share it. But yeah, I mean, it's wild because, oh, now 112 years later, we still have the same fears about what next musical technology is coming around the corner. Beautiful, Nate. So this is old hat.

Am I redeemed? I'm intrigued. Super redeemed. Yes. So thus it ever was, right? As much as it may terrify me, too, to think of these AI composers taking over, it's probably more of the same. I don't know. I want to ask Terran a bit more about this because we don't want to use a heuristic to lead to a false conclusion. Terran, earlier in the show, you suggested, hey, musicians don't worry about it. AIs are pretty bad at making a good composition for the next couple of years, maybe a decade or so.

Where do you see this technology going from the research that you've done into it? Okay. This has been a great show, guys. I've got to say. And we've gone through a lot. of really interesting topics. And I think the article that you just shared hits the nail on the head in terms of how most people view AI with this sort of terror that John expressed in early 1900s with mechanical music,

which of course is now a staple of probably a lot of musicians in their work. But I'm curious to ask the question of what will be birthed as a result of these new tools entering the marketplace and how that will create

a whole new area of need that we can't even yet imagine or contemplate in the same way that being a YouTube video creator was not even possible 15 years ago because YouTube didn't exist and the idea of digital video 25 years ago was actually impossible to fathom and we've created all of these new areas of artistry where creatives are needed in different capacities and I think the same will be the case with music. I actually do think that AI will get to a point of creating fantastic music probably much sooner than 10 years from now,

just based on the relative exponential rate of technology growth and the money that's now being invested in this area. But it's what we choose to do with it as humans and as artists. We always find really interesting ways to evolve with the times. And so I was actually just talking today with a guy who runs a music. publishing company and he called me and he was both concerned and excited about the possibilities. And he said, I don't know what to tell some of my young clients about whether they should be entering the world of musical composition. And I'm like, I don't really have the right answer to that. But I would imagine that using these types of tools that will have all kinds

of new games to play and new types of entertainment to enjoy, I can imagine that haptics and that augmented reality and that these new, these other new technologies that are just coming onto the marketplace could interact with music in ways that we can't even yet imagine to create experiences for people. And so people who have like a real, not just ear and passion for music, but experience the world through music could probably dive into that in some insane way that we just, we can't comprehend yet. And so I'm excited about that. But the reality is that, yeah, I would say many of us may not be looking at music or composing music the same way in 25 years. And that's okay. Wow. Mind blown.

I also would add that you're going to have like a whole group of young people, right, that have never had traditional music education or backgrounds because we have so little funding for it, not just in the states, but around the world. And if you're a kid who wants to make a song and your first entry point is through these tools and you possibly know how to code, maybe you do, maybe you don't, but you're figuring out how to create with a different language than one that is used in music education classrooms,

you are inevitably going to, there is like, I don't even know what to think about what they'll do with the tools, but it might be something totally different than what we would do, given our traditional music education, because there's a frame and a background that's really hard. It's a box in many ways that's hard to jump out of. But if you're not given the rules, then you just create something new.

It's kind of like, I would imagine electronic music and hip hop how, like, early days of hip hop, a lot of traditional musicians are like, this is crap, this isn't music. And hip hop just like did their thing. Oops, wrong prediction. Yeah, they did their thing. And they've really transcended into multiple genres. as well as electronic music. So early tool adopters

just think differently. Can I make a prediction then? This is ridiculous. But it makes me think of a really bad episode of Family Guy, which encompasses, I think, all episodes of Family Guy that I haven't watched since I was a teenager. But there was an episode in which Peter Griffin, the lead character, father figure,

somehow, I don't know if he asks a genie, a wish or something. He ends up having an orchestrated theme song following every single action that he does. And it's wildly entertaining and becomes quickly very obnoxious because the theme music never goes away. So perhaps in the future, I can conduct and compose my own theme music at all times that perfectly represents my state of mind, my mood, whatever's going on.

So there's my ridiculous prediction. I love that prediction. And I actually agree that that will probably happen. Yeah, weird world. Yeah, I have my own sort of internal symphony happening sometimes. So if I was able to actually create that on the fly and then share that with others, while also throwing in some augmented reality graphics and touch haptics at the same time, that would be pretty cool. That's awesome. It's like my own romantic comedy following me 24-7.

I feel like this has been an extraordinary, I guess, remaking of all of my thoughts about artificial intelligence and music. Because I feel like your music and your compositional process has both, help me really understand the limitations in the context of what these tools are being used for today and certainly getting rid of all the sort of marketing hype that it's going to ruin our lives. And at the same time, you've given me a great optimistic view of all sorts of new creative forms that I can't even imagine. So thank you for all of that. And I want to know where can we find your music? When's it coming out?

What's going on? Thank you so much. Well, all my music goes up on iTunes and Spotify and all the usual players. you can find the official music videos for the songs on my YouTube channel, which is YouTube.com slash Terran, T-A-R-Y-N. And, yeah, I'm on all the regular social media channels. My album comes out this September. But like I said, I've got a couple more singles that I'm releasing until then.

So you can just subscribe on YouTube to get those as they come. Fun stuff. Thank you so much, Terran. Yeah. Thank you. Switched-on Pop was produced by me, Charlie Harding. And me, Nate Sloan. Thank you so much to Terran Southern for just. joining us today on the show. Our editor and mix engineer is Bill Lance. Our designer is Luke Harris,

and we are a proud member of the Panoply Network. You can find more episodes of Switched on Pop on the Apple Podcast app, Spotify, whatever player you prefer, or check out our website, switchedonpop.com. We'll be back again in two weeks with more episodes. And until then, thanks for listening. Thanks for listening.