Can AI music be good? (Part 1)

0:00I'm Annie, Noah, Devon. And this is no such thing, the show where we settle our dumb arguments and yours by actually doing the research. On today's episode, can AI music be good? No such thing, no such thing, no such thing, no such thing, no such thing, no such thing, no such thing. No, I'm assuming that was AI. I'll explain after the break. This is an I Heart Podcast. Guaranteed Human. On Big Lives, we take a single cultural icon, and we pull apart the story behind the

1:08image, discovering forgotten interviews that change exactly how we see these giants of our culture. We're here for the messy, the brilliant, the human version of our heroes. I'm Emmanuel Jochi, and this is Big Lives. Listen to Big Lives on the I Heart Radio app Apple Podcast or wherever you get your podcasts. Is there an epidemic against literature? More specifically, Black Literature. I'm Jack Weiss-Peace Thomas, and on this season of the Black Lits podcast, we're outside asking why. They don't want a whole bunch of little elevators, Malcolm X's and Martin Luther King's on the line. And when you look at a lot of different cultures, they're like, we need to know where you came from. Considering this is Black Reading Month, the only real question is, what are you reading next? Listen to Black Lit from the Black Effect Podcast Network on our I Heart Radio app Apple Podcast or wherever you get your podcasts. What up this Jay Valentine co-host of the R&B Money Podcast alongside my brother, R&B Legend, Tank. Together, we've built a space for artists, executives, and fans of R&B music to reminisce

2:11and tell the real stories behind the music. I have no money, I was hood famous. They think you're going to get everybody out. We lived in a project. The world love came out. They hear them music and we had nothing. From Smokey Robinson to Mary J. Blige, B2K to Jill Scott. If you're a super dedicated R&B fan, this is where you belong. Listen to R&B Money from the Black Effect Podcast Network on the I Heart Radio app Apple Podcast or wherever you get your podcasts. All right, fellas, I have some explaining to do. What was that? As you might have guessed, based on the topic of this episode, the alternative, no such thing theme song you heard before the break was generated by AI. As a journalist, I thought that if we're going to do an episode about this, I should try using the program that most people use when they make AI-generated music. That program is called Soono. What exactly did that entail? I uploaded the no such thing theme song that I actually produced without AI.

3:11If you're on hand. It was my own brain and creativity. I told Soono to make a Motown version of it. I just literally prompted the program. I said, here's my song. Can you make this sound like a Motown song? That's the only prompt you gave it? Yeah, one line. What did you think of the song? It was silly. It reminded me of... We're back in a day where there were those... This is not an audio comparison, but we had those little digital postcards almost that you could send your friends where you could put their faces on like all sorts of things like that. Where it's silly. It's kind of a template and just add. Exactly. It reminded me of that of just sort of like, okay, you doing this for this episode. Oh, it's a silly, sort of thing. It was catchy. I don't think it's something I'm like, oh, we need to replace the theme song. But did you think it was catchy? Something that you might listen to without the context of me recreating our theme song?

4:15No, absolutely not. No, it was a little... And granted, I love our theme song, man. It's a great work. But this being based on that, you can really tell it's like copying, pasting that same no such thing. Yeah. Which is different when it's a spoken word of someone who won't be named. Singers. Yes. The first couple of times it was like, okay, this sounds good. If it ended there, I'd be like, that was nice. Yeah. If we wanted that style, that could work. Listening to it for however long, full 20 seconds, I was like, okay, I can tell this is the exact same piece over and over with no variation whatsoever. Yeah. And as an amateur musician, music producer, that's how I felt. Like when I heard at first, on the first couple of lessons, I was like, impressed that it could come up with this. After like 20 lessons, I was like, okay, I can hear where it's getting its source and ideas from. Which makes sense. I literally uploaded the song. It might have sounded completely different if I prompted it without uploading a song.

5:17And I think the, you know, AI music defenders would say, of course, it didn't veer too far from your song because you only gave it one line. I think those guys are really doing a lot of typing when they're recreating their music. But just to give you an example of how easy this is, I was blown away actually at how easy it was to turn this around. I'm going to play you some other genres. As I said, I uploaded my song and it created different genres of music with it. So Connor, can you play the one that's called UK Garage? I guess my wife would say, but you're on. Well done. You're rich. So yeah, that was the UK Garage version. The one on. The one on. You're rich.

6:17Yeah. It's really interesting. Like if I were to put South American hip hop or whatever, it would take the drums that we all know and well, the, and I think that's just what it's doing for every genre. Okay, so I prompted it to create a classical version of the no such thing theme, which it does at first and then it kind of changes. No such thing. That's beautiful. Damn. Is this the Odyssey? Oh. Not really classical music, but so that's you just put in classical and it did this. Yeah, I was, my prompts were almost one or two or three words only. That didn't really write full sentences. That's interesting. Yeah. I think it's like, it's trying to do classical music, but it knows that the song that I uploaded is has drums and percussion. Yeah, so trying to. There are like settings that you can't access unless you pay, which are like how closely

7:19do you want it to resemble your original song? What's the weirdness level they have? There's like whether you want the vocals to be a man or a woman. So there's a lot of like fine tuning you can do. The next one I plugged in I think is called what it's a genre I call Christian Kanye. No such thing as a dumb question. No such thing as a dumb question. No such thing as a dumb question. No such thing as a dumb question. No such thing as a dumb question. No such thing as a dumb question. The lead vocal on this is better than the others I think. Yeah. As far as just a little more character. It's fake Charlie Wilson. Yeah, I think it's literally Charlie Wilson. Which should be illegal I think. He has a right to say. Yeah, definitely sounds like it's being trained on Charlie or someone similar. Some people, some artists, AI music artists, not fake ones, but real guys who make AI music

8:21can do this in ways that are much better than what I just played you. So this what I just played you was like pretty remedial. I gave Suno one sentence prompts and it created these things. I could understand why some people might be like, oh, this is catchy enough for us to use it as whatever commercial music. But I want to play some music that so called AI music artists have put together. And it's music that kind of I have an uneasy feeling about because it's good. So what I'm going to do here is play you the original song and then I'm going to play you the version a specific artist called almost real. He takes the original song and he creates Motown versions of those songs. So first is many men by 50s and off to get rich and die trying. I'm trying to be what I'm destined to be and he just trying to take my life away. I put a hold on the nigga for fucking with me. Okay, well you guys know the song.

9:22Yeah, yeah. Okay, Connor, now you can move to the AI remix. So this is supposedly a 1950s Motown version of many men. Man and man wish death upon me blood in my eye dog and I can't see. I'm trying to be what I'm destined to be and bones trying to take my life away. I put a hole in a dummy for messing with me. My back on the one. So the AI redoing of the original song here seems to be more involved than what I was doing. I don't know exactly how he does it. I reached out to almost real multiple times. He did not respond. I imagine it's a similar process, but one that's just way more nuanced, especially because when you listen to that song, you can hear like you can kind of hear the separation of the different instruments and stuff. I wonder how much composition he's doing before he gets the AI vocals involved. But I just wanted to hear your guys' thoughts on the difference between what you just heard

10:24now and like what I was showing you earlier. This feels like a, the many men version feels like something you would see on like a YouTube parody. You know, where it's like, okay, obviously this person has some sort of musical talent. They're working from a thing that already exists, so they're not making it from scratch. So it's, you know, I don't love where it is, but you know, like that version of a thing. To me, it feels like more like, okay, this is a person who, if you had given them the resources and studio time, they could have worked with a team to make this thing. Yeah. But because they don't have that, they're using the computer in a way to, you know, create orchestras and instead of hiring a vocalist to do this thing, using the AI prompts to either take a vocal and alter it in this way or create the vocalist scratch based on a reference. Yeah, exactly.

11:25Almost real. He had like dozens of these songs on his various channels, but they were copyrights stricken to hell, obviously. The argument that he made in this very long post that he had on Reddit was that these are just covers. These are AI covers of music and covers are famously a lot like you don't need to get express permission. I don't think to do a cover. Music law is so wonky. There's like, you could be doing a cover of the composition that they did in the studio or you could be doing a cover of the master that came out. So it's like, there's all these different arguments. Basically, these, what the labels are saying is that this is not a cover, actually. This is a new work that just rips off our original song. And so that's why these things don't exist on the internet anymore, but I'll play you another example of another one that I thought was admittedly catchy. So first is the original, it's Kid Cuddy Day and Night.

12:28Okay, we can move to the next one. This like is catchy to me. It feels good. It sounds nice. It is obviously ripping off of Kid Cuddy's voice, but it made me ask myself like eventually will we be able to really tell the difference. Does it matter if AI was used in making the music if you feel like it's good music? For me, just for my music production experience, using Suno was a novelty to me. Like I was interested in playing with it as essentially a new technology the same way like the Nintendo Wii came out and all of a sudden we're playing with motion control

13:32and it just felt new and fresh. But then eventually we're like, this is kind of stupid. Let's look back to the regular controllers. So it felt like a novel to me that was fun to play with. But then after a couple hours I realized like, oh, this doesn't give me the same fulfillment as for example making the no such thing theme song. When I finished that I felt like I had really accomplished something and that I used my creative abilities to produce something that people like. And with the Suno thing, the end product was just something that I thought was kind of cool but like I would forget about eventually. But before we go to break, I want to know your guy's opinion like does it matter if AI was used if the song is good? You know, AI Cutie is what he thinks he sounds like. I'm just kidding but in that I'll say what's kind of good about the original version is he's not a great singer and it has a lot of character that you're not going to be able

14:32to make a little fake at this point. Like I'm sure you can add more prompts and get that voice, the fake guys voice a little to match a little bit better and have more imperfections. But that's kind of what makes that song special. And a lot of music, I don't listen to music I like because it's the most perfect music. In fact, I often like things that have more of an edge or something else to it. It's really kind of grab onto and it makes it feel more real. And if everything was just like punched in and came out exactly how everyone wanted it to be, even if maybe the artist would have preferred it to be that way originally. In time, most of what I really like has something a little bit imperfect about it. So I don't think it's impossible to enjoy music made by AI or to use it as a piece of something but it sounds too polished. Yeah, I think that's overall my biggest concern. Dev and I saw you bob in your head during the kid. Yeah, I was snapping. Yeah, what do you think? No, no. I would never listen to that song if that was the only version of the song.

15:34Yeah, yeah, yeah. And yeah, like, you know, it sounds nice. A lot of music sounds like that. It's good for like a Walmart. Yeah, sort of. Exactly. But to know this point, I think this is exactly what a lot of people who are utilizing AI don't understand about how music connects with people is that we like to imperfections. That's the same reason I don't really love when like great singers cover songs by not great singers. You know, like there's one of my favorite songs ever. It's Kanye street lights and there's a Daniel Caesar version and Daniel Caesar definitely is a better singer than Kanye West. But I don't like the Daniel Caesar version because it's saying so well that it kind of, you know, loses a bit of imperfection of Kanye's voice. Yeah. And it feels a little bit like, this is not quite what this song should sound like. And it doesn't connect with me in the same way. And it's cute, right? I watch the YouTube video. I'm like, oh, that's funny. And that's another take on the thing. But I'm never listening to that other version of it. But you know, maybe this is the early stages of this.

16:39And I actually just over the last couple of weeks have had the reckoning with this myself because I start listening to this artist who's not all that new, but he's new to me called 1010 Banja. He has his new EP. It was out that like another artist that I really like posted about and I started listening to it. And it was like one of these projects that like I just had on repeat and I was like, wow, it's been a while since I felt like this excited about an artist that's new to me. And then I followed him on Instagram and I started seeing people posting in the comments, like don't listen to this. This is AI. Oh. And I was like, what do you mean? This is AI. It's like the artist who recommended him as like someone who I deeply respect. So I started looking into it. And he went through a process where he created the song, fed it into the same program that you use. Kind of perfect. And he used to kind of a different way that these people are using it and that like he

17:40basically was trying to get the AI to hallucinate and do weird, crazy things. And then he took that version of the song and gave it a bunch of prompts. But then that ultimately, the combination of what he drew originally, spitting into this thing, putting that back into Aebelton and playing around with it, that became the version of the song that's out. Quick side note. If you want to hear Devon confront this artist he likes, 1010 Benjo, check back in on the feed Friday for part two of this episode now back to the episode. You think it's worth less because AI touched it. Yes. To some degree, the instances where artists are using AI, it feels like a sort of laziness. Like you talked about like there's obviously something much more to you creating the thing from scratch versus putting it a prompt and it just playing it. It's interesting. People like Kanye West who are very open about using AI in their music.

18:44There are artists like Charlie Puth who has been adamant for years that AI is going to be like it's inevitable and we should just embrace it. There's the new Tiger album, which he's very open about the fact that he used generative AI and it also got a 0.0 review on pitch fork, which I thought was pretty funny. There's also all these allegations that play boy Cardi is using AI in his music, although his team flatly denies these allegations. So it's not necessarily the fact that AI is this kind of experimental field in the music industry right now. It's pretty integrated into music production workflows to an extent that I don't think people are very aware of. And there's really two ways to look at this I think. One is someone who's doing generative AI like you're prompting something to make a song versus some artists and audio engineers and producers who are using AI in their workflows

19:50like mixing, mastering that kind of stuff. And I think it would be helpful for us to understand exactly how that stuff goes down. And so after the break we're going to be talking to Charlie Harding. He's the co-host of the Switched On Pop podcast over at Box and he'll be schooling us about this. Hey everyone, it's the Jonas Brothers with the Hey Jonas podcast. We've been catching up with some great friends Paul Rod, Michael Bubley, Seth Meyers, Nile Horan, Jake Shane, and making some new friends while getting into what's really like to have sisters like Alex National. She likes to control everyone and tell everyone to go. Okay well control is a harsh word. I would say that's a harsh word. I am more outspoken. Growing up a lot of the times because she was so quiet I would speak for like the both of us. Now I'm working on like thinking a mom is really not a thing before I speak. And Joey King. You know if your like mom is yelling at your sibling and even if you're annoyed with

20:54your sibling you just like won't let that slide. It's like you only can yell at your sibling. I'm like you can't do that here. That is not what I want about her like. I just want to talk about her. Plus we've been taking your calls listening to your voice mails and giving advice or at least trying to. So if you've got catching up to do now is the perfect time. Listen to Hey Jonas on the I Heart Radio app Apple podcast or wherever you get your podcast. Is there an epidemic against literature? More specifically black literature. Our books are being banned, scanned by AI and destroyed. I'm Jack West peace Thomas and on this season of the Black Lit podcast we're outside asking some of the custodians of our stories teachers and our modern day writers why they don't want a whole bunch of little elevators. Malcolm Axes and Martin Luther King's on the route. They want to build children who have no empathy. That's really the ultimate goal. We want books that don't make white kids feel bad. So is the war on literacy and affirmation or our ability to connect?

21:55Well people to know how sexy reading is and how fun they can be and not always associated with education. They could be in a state. They could be entertainment and that is the beauty of reading. It can be all of those things. And considering this is black reading month, the only real question is what are you reading next? Listen to Black Lit from the Black Effect podcast network on our I Heart Radio app Apple podcast or wherever you get your podcast. Hey, what's up? I'm Tony Gonzalez and I'm October Gonzalez, but you can call me Toby. We want you to join us on our new podcast No Say Forwards with October and Tony Gonzalez. We get real about marriage, parenthood, sex, divorce and all the nitty gritty stuff we all think about, but just don't say I hate that job. You're probably just going to meet with some rich guy and he's going to whisk you away and I'll never see you again. Oh, this is all I would say a deal is actor. D? I don't even know there's this thing is D. No, no, no, I mean like I didn't get on the sea list like I'm not even average.

22:57I'm not even passing. I'm failing. Listen to No Say Forwards with October and Tony Gonzalez, part of Michael tour podcast network available on the I Heart Radio app Apple podcast or wherever you get your podcast. All right, we are back. I'm Annie Noah Devon and we're joined by Charlie Harding, who's the co host of the Switched on Pop podcast over at Vox. Charlie, thanks for joining us. Thanks for having me. So just now we were talking about AI being used in music, some artists being very open about the fact that they're using AI in their music. I just wanted to get a layout from you, just how much AI is being used in music today. Do you think is it like a very prominent thing or is it still kind of new? It just depends on who you ask. You know, we live in this moment where there are so many people whose job it is to promote this stuff. And so I think we're hearing wildly different reports from the people whose job it is to share this material because they might be sponsored by Suno.

23:58You have industry people who are working on deals with these generative AI companies who want to tell you that, oh yeah, everybody making music are using generative AI. I have seen in my own reporting that there's widespread use of AI tools in Nashville, like the last place you'd expect. Dirt roads, blue jeans, they're making them with AI these days. But exactly what people are doing, it's not like the whole songs are being created with AI and what you're hearing on the radio right now are AI generated songs. I think there is both a widely overhyped sense that everybody's doing this, but also maybe there are small parts of people's workflows that are starting to include artificial intelligence. Yeah, I have friends who are musicians who have told me about using AI in their workflow process versus just kind of prompting a program to create a song for them. Things like mixing and mastering, you're an audio engineer, music producer yourself to use any of it and to kind of what extent to use it. So many smart companies are not advertising AI, right? Because AI is kind of becoming a bad word.

25:00So I probably couldn't tell you all the tools that I use in the mixing and mastering process that may have been trained on a neural net architecture, right, which is technically artificial intelligence, but is a substantively different technology than generative AI. So there's a lot of mixing plugins that might do some EQ-sweetening, right, that technically are an AI-type tool. These kinds of plugins and the mixing and mastering process have become absolutely ubiquitous, but they're often used alongside traditional tools. They're often just like better, faster ways to get a slightly nicer mix. Gotcha, gotcha. Earlier I played for Devon and Noah, these kind of AI covers, quote-unquote covers because there's a debate about whether their covers are not. Are you aware of those and how exactly are they being made? I mean, they're cute gimmicks, right, at best. There are generative AI models like Suno that will let you cover your own music. In fact, this is how Nashville is broadly using Suno.

26:00You can upload a really cheap sounding demo, something you've just done on an acoustic guitar and you're singing along and you could say, hey, make it snappy, make good. And they'll spit out a so-called cover or remix of your own song in the style of a fully produced Nashville country track. That's being done really as a way of getting faster demos so people can sort of get a sense of what would this sound like if it were fully produced. Those songs are not then getting released. They're then getting sent to publishers. Those publishers are pitching them. Those songs are getting re-recorded in a studio. You can't do that technique with copyrighted material. The funny thing about a lot of these covers is that they're actually pretty hard to make. You have to first strip the vocal from the instrumental. That's not hard. You then have to basically find all of these sort of legal gray area voice models that people have trained off of. Other people's likeness and they're available on the internet. If you do a lot of deep searching, you then have to run your own local open-weight AI

27:03model to generate the voice in another person's voice. You then have to use another open-weight model to create music, pair them together. People are putting a lot of work into making these covers, which makes me wonder why they're not just putting the work into learn how to play an instrument and cover the song. Another of the examples was Kid Cuddy's Day and Night, which was remixed into a Motown's version. This was, I think, before Suno stopped lighting people upload copyrighted material because this character online, his name is almost real. He had produced all of these songs and then all of the music labels made him take them down because they are not technically covers according to the music labels there. Instead, new works that are ripping off the original work. Could you kind of lay out the difference there? What exactly is a cover versus what this AI recreation is doing to the original songs? Well, I mean, one thing we can all know is that the US copyright office currently does

28:03not allow a generate of AI music to receive copyright protection. Obviously, the underlying composition does if it is stolen from somewhere else. When someone creates an AI version of Johnny Be Good, that recording cannot be copyrighted, but still Johnny Be Good is still a copyrighted work. This song is in lyrics and so on. This is something that the music industry is actually, unfortunately, I think, very excited about. They see that there is this desire amongst fans to interact and play with their music to create their own covers. This thing that we're not supposed to be doing currently, all the major labels are working with AI companies to create legal forms that you can remix pre-existing music by the musicians with their permission and they then get credit. This is sort of like an all behind the scenes legal mechanization that we still have yet to see exactly how it's going to play out. But the difference between an AI cover and a human cover, I don't know. It both sound like a cover.

29:05Again, one is copyrightable, one is not. If we all decided to start a band and make a cover, we all have the right to cover any song we want with a mandatory license that is given out to us. We can all cover the song. We can play it on YouTube. We can do something really creative and we can own that master recording. We can put it on streaming services, but you can't do that if you cover it with AI. Even though, like I said, the music industry is really excited for it to help you do so in some future fan tool that they would like to imagine is going to get a lot of use. Just going off of that, like where do you see all of this going? I mean, are we going to start turning on the radio and hearing fully AI generated music on the top 40 stations? I definitely do not have a crystal ball, but I just got back from the Newport folk festival, which is one of our old running famous summer festivals. It's a place where there is a lot of protest music. It's a place where people make big statements. And it's a place where you don't hear any backing tracks. You hear completely live music. There is no AI. The fans go nuts.

30:06We listen to music. We write music because we want to have a human connection. Invariably, someone is going to trick us. And there will be a top 10 song that is primarily made with a genre of AI. And at some point, the headlines are going to say, we're all cooked. And I just don't think so whatsoever. Music is about the transference of emotional compression, the things that I've experienced in my life. I'm going to go down to a three minute little song and I'm going to give that to you to experience. If you compress that fan relationship in the creative process into some generative format that has no actual experience behind it, we're just selling smoke and mirrors. Yeah, I was encouraged to hear Devon and Noah. I played them the songs I was talking about earlier, these Motown remixes. And they were very easily able to tell me why they thought the original recordings were much more impressive. I generally feel like I teach songwriting at Berkeley and in my U. And I constantly teach my students that it's important that we make sounds that are

31:08honest. Doesn't mean they have to be good or bad, but they should be what they are. When you're trying to fake something, people can often tell when it's just like it doesn't sound quite right. That includes like using, I don't know, a poorly sampled electric guitar on a keyboard. Will sound wrong. But what also sounds very cool is if you go back to like really Destiny's Child, there was a lot of really poorly recorded acoustic guitar samples that are played not to sound like a guitar, but this like other worldly weird version of a guitar that also is honest. Yeah. Perhaps there's some like interesting place of creativity there. Using Suno's models that have been trained entirely off of copyrighted music and the goal isn't to create your own sound, but often to sound more alike something else. I'm not excited by that. Yeah. That was going to be my next question is just like, what is it exactly about Suno that upsets you as an artist? Ooh. How long do we have?

32:10I mean, there's a whole litany of things that I would say I know and don't know, which is like exactly how did they train their music. It has been largely reported that they did so by scraping the entire corpus of music that they could find on the internet, which means music I composed is probably used to train their model. I'm not thrilled about that. I also know that anthropic and open AI stole my book in order to train their models. There's something that is a little bit different in the output of large language models, outputting text than these tools that are opening audio. For me, that is if I went and asked Claude to give me my book, he would do a terrible job. It's actually pretty bad. It has such its own style of writing. I don't think it's particularly good at replacing what I have written. When we're talking about music, whenever I use Suno and to be clear, I have used Suno to try all the tools, see what it does, understand it as a journalist. I haven't used it in my own creative practice. But whenever I use it, it just makes my song sound like the most generic version of someone else's sound.

33:11What I get out of these tools is that it is all about sounding like someone else, not developing your own voice. There's a whole litany of other issues. I'm sure of, are they stealing energy from small communities that we've seen and all of these issues with data centers that are happening? I'm not deep on the reporting on how exactly these tools are using what their infrastructure is, etc. I'm mostly concerned about it from a point of creativity. I don't like that we're trying to sound like other people. Frankly, that's not what I'm trying to do with my own creative practice, nor am I trying to teach that to my students. In preparation for this episode, I was on Suno a bunch yesterday. Yeah. Yeah, I had the same feeling of like, it was kind of fun to play with as like this new thing that I hadn't played with yet. But at the end of it, not feeling the same kind of fulfillment that I would feel after I create, for example, like the theme song to our podcast, the feeling of like achieving something that someone might like. And so, yeah, my like kind of take when I was done using Suno's, like I'll probably never

34:12use this again. It was cool for like a half an hour, but it was, it was a novelty that I just, like I got through immediately. Making a song requires thousands and thousands of small choices. Like when you get to the mix stage alone, you got a hundred tracks. Those hundred tracks are all getting different processing, right? That's not even thinking about the beginning part of like, how do you go from music to words? How do you pair phonetics with actual rhymes? What words do you choose at each, in a given moment, what are you trying to say? How do you express what you're trying to say and pair it with music? Like every one of those choices, it's like cooking a great meal. Like every choice that you make when you're cooking adds up to the final dish. And I just don't think that there's a cheat code. You're like, oh, there's my song. Done. It perfectly represented what I was trying to communicate. I just don't believe that that is possible because the process of writing is the process of discovery of what you want to say and what it means. And it just is a fundamentally different product when it's like the cheap, instantaneous

35:15version versus the thing that you actually meant to make. Yeah. So we're all on the same page about like, Suno essentially. Like it just feels kind of lazy and tacky. One thing I struggled with is like earlier I mentioned I have friends who use AI processes to like mix and master their music. Where do you square that in terms of like the creativity? Is that also kind of lazy? Is that something that is going to become more and more common? And I don't mean tools within the DAW that are technically AI. I mean like they'll upload the stems to like a website and then it'll just kind of spit out something for you in different presets. How do you square that like the workflow of having AI? I mean for me, anyway that I'm using technology ought to be to enhance my creativity or better to protect my creativity. Like I am all for my DAW building a tool that is like mixed prep my song. So prep is so boring. That is like relable everything. Put it in the right order, color code it, put the right icon on it. Part of that you know sometimes doing the work of organizing helps us understand our material

36:16better. Sometimes we don't want to skip those things. So I do want tools that will help me be more creative by protecting my time, getting rid of the things that are busy work. I hate pushing pixels. I hate having to interact with my computer. I want AI tools or whatever kind of tools are necessary to get rid of all that stuff. The idea of drag and drop your song and it is mixed. I mean the mixing phase is also a creative act. It's definitely a bit more of a service business, but a great mixer what they're doing isn't mixing to the idea of what a pop song sounds like or hip hop song sounds like. They are hearing what is this song communicating and what choices need to be made in order to enhance the emotion of the song. And again, I don't believe at this moment that AI is at all capable of comprehending the emotional impact of a given moment of sound. And even if it ever is, I still like that music is this human interaction from the full act of creation and production to listening and experiencing.

37:19Yeah, I've heard so many of those rumored stories about how the Kanye West song we're hearing on the radio is like some audio engineers 1000th mix of the different mix of that song. The nine of us all right now that every little choice down to the like one decibel of how much bass is happening in a particular moment is affecting how I'm experiencing that song. Now, is the average listener going to hear that? No, in the same way that the average listener doesn't understand what's happening behind the scenes in the kitchen, right? Like they don't know when they're when you're going to a great restaurant exactly how your meal is made, but it is all of the craft and experience of that maker who is able to generate this beautiful thing that we get to experience, which is art. And so, you know, all those choices matter and some of them are more subtle, but they all add up. I guess we talked a little bit about the fact that artists are now, like you've been saying, using AI is kind of a shortcut. You know, we talked about the idea of like playboy, car, or Kanye using AI, you know, for

38:20vocals. But it seems like now it's really up to defense to try to decipher between, okay, like that kind of sounds like AI that doesn't. What are some of the things that you look out for that are like tells for you that like, okay, you know, these maybe these vocals aren't a real person singing. So it totally depends on if we're trying to listen to a fully generated AI song versus if a portion of it is generated. Yeah. You know, this requires us to put on a really different set of years. We are predominantly a visual culture. I think goodness people listen to podcasts and still like using our ears and attention in that kind of way. But when we're listening to music, I think it would be really helpful to understand thinking almost in multiple dimensions of depth. What is in the foreground? What is in the mid ground? What is in the background? And elements that are in the foreground that are generated AI, we can almost always tell. And it's probably a place where I feel more up in arms about why are you replacing a full vocal, the thing that is the most important with AI. If you have some really far off background element that somehow use generate AI, it's

39:21so reverbed out, it's low in the mix. I can't know, I'm not going to be able to hear that. But what I am listening for is this sort of like grainy quality. It's kind of like if you take a picture and you get like the really bad low quality JPEG version of it, you can see the pixelation. You can hear sort of a pixelation of audio we call artifacting. It's usually more in the high end of the material. And this is largely because the music that the AI's were trained off of were lower quality MP3s. And so what they're generating back, there is this sort of like lower quality grainyness to it. And so that's what I'm listening for that is almost always a tell on top of just like weird things that shouldn't happen. Like sometimes we hear like a reverb that will cut off suddenly in a weird way that is unnatural. Sometimes like something will be in the front of the mix and it's sort of weirdly fade back into the mix. So there's just like these weird hallucinations that occasionally happen. But pay attention to the high end if you hear some fizziness that might be a sign of AI. But frankly, there's so much music in the world.

40:23And there's a lot of bad music. Go listen to good music. You know, like stay away from the auto play, go find the things, talk to your friends. That's we can avoid this stuff. Charlie Puth has been tapped as like chief music officer I think it's called at a company called Moisés. Darry away of what Moisés is. Is it different than Suno and any meaningful way? I haven't checked in on Moisés in recent months. I have known it to be a STEM separation technology, which is I think categorically different than a generative AI. If you're using an NPC or if you're using a logic, if you're using Ableton, a lot of these tools have STEM separation built into them. They're technically using a form of a neural net architecture to be able to pull, separate bass, piano, guitar, drums, vocals from each other. I think that is a different kind of thing. And largely being used by DJs, remix people and I think is more broadly accepted. It doesn't have the sort of same taboo as a fully generated song. Yeah, I remember back in the day when I would want to get like a Nas verse from the song.

41:26For some technology. We're going way back. I would use audacity to like, you'd like copy the song on top of itself. And then it could strip the... So like for it doesn't really bother me if people are using Moisés to strip vocals when they were doing that the hard way before Moisés came out, if that makes sense. It feels like something you were going to do anyways, but now they've got this technology that gets rid of the busy work, the tediousness of it. It's also imperfect when you're ripping a vocal out from the entire instrumental that vocal has all kinds of processing. When we master a song, when we mix a song, we're working on gluing elements together. You can't undo the glue. Like that glue is still there. There is no googon for the sort of strange artifact you hear in those separated vocals. And frankly, so much amazing music has been built off of DJ's finding creative ways to isolate parts of the music. We might use a filter to cut off a lot of the high end, so you kind of just hear the kick drum. But still, other things are poking through. And sometimes that way of getting stems can create some really weird and original work.

42:31And that's how so much great, how some music was developed, so much great hip hop was developed by people trying to sort of force isolation that they couldn't completely get. And we're thinking of all the great drum breaks of history. Sometimes there's some weird thing in the background, and we like that stuff. One of the central questions we want to answer in this episode is a little bit more philosophical. And so I want to see if I get your take on this. If a song is emotionally moving or resonates with you, does it matter that it was created by AI? Why or why not? I can't please people's emotions. That's a ridiculous thing to ask. I do think that we have responsibility, especially with those of us behind microphones, to help educate about what these technologies are, how they're being used, that may impact then how people hear things. What is where I feel like I can insert myself in the discourse? But if you heard BBL Drizzy and you had a good time, I can't be upset.

43:31Yeah, you did have a good time because it was a funny novelty song where mustard used, I believe, one of these AI tools to create what sounded like an old throwback thing. And that song, again, novelty, right? We're not still listening to that track. It was part of a hip hop discourse. A great rap battle is one in the middle of the whole thing. It wasn't the song that was the biggest. So anyway, you're going to be experiencing AI in your life all over the place. And I can't tell you how you should be feeling, especially if that is your genuine reaction. But I feel like we can, hopefully, educate people on the sort of ethics and responsibilities we might have in using and consuming these tools. Well, Charlie, thank you so much for joining us. This has been an illuminating conversation. And no such thing will be right back after we pay some bills. Hey, everyone. It's the Jonas Brothers with the Hey Jonas podcast. We've been catching up with some great friends, Paul Rod, Michael Bubley, Seth Meyers,

44:31Nile Horan, Jake Shane, and making some new friends while getting into what's really like to have sisters, like Alex National. She likes to control everyone and tell everyone to do that. Okay, well, control is a harsh word. I would say that's a harsh word. I am more outspoken, growing up a lot of the times because she was so quiet. I would speak for like the both of us. Now I'm working on like thinking of mom. It's a really nice thing. Not completely. And I'm not thinking before I speak. And Joey King. You know, if your like mom is yelling at your sibling and even if you're annoyed with your sibling, you just like won't let that slide. It's like, you only can yell at your sibling. I'm like, you can't do that here. That is not what I want about her like that. I just want to talk about her. Plus we've been taking your calls, listening to your voice mail and giving advice or at least trying to. So if you've got catching up to do now, it's the perfect time. Listen to Hey Jonas on the I Heart Radio app Apple podcast or wherever you get your podcast. Is there an epidemic against literature? More specifically, Black literature. Our books are being banned, scanned by AI and destroyed.

45:34I'm Jack Weiss-Peace Thomas and on this season of the Black Lit podcast, we're outside asking some of the custodians of our stories, teachers and our modern day writers why. They don't want a whole bunch of little elevators, my comics is in Martin Luther King's run around. They want to build children who have no empathy. That's really the ultimate goal. We want books that don't make white kids feel bad. So is the war on literacy and information or our ability to connect? Well, people with a know how sexy reading is and how fun they could be and not always associated with education. It could be in a state, it could be entertainment and that is the beauty of reading. It can be all of those things. And considering this is Black Reading Month, the only real question is, what are you reading next? Listen to Black Lit from the Black Effect podcast network on our I Heart Radio app Apple podcast or wherever you get your podcast. Hey, what's up? I'm Tony Gonzalez. And I'm October Gonzalez.

46:35But you can call me Toby. We want you to join us on our new podcast No Say Forts with October in Tony Gonzalez. We get real about marriage, parenthood, sex, divorce and all the nitty gritty stuff we all think about. But just don't say I love it. I hate that job. You're probably just going to meet with some rich guy and he's going to whisk you away and I'll never see you again. Oh, this is all I want. Would you say a dealers actor? D? I don't even know there's this thing is D. No, no, no, no. I mean, I didn't even get on the sea list. Like I'm not even able to have. I'm not even passing. I'm failing. Listen to No Say Forts with October in Tony Gonzalez, part of Michael tour podcast network available on the I Heart Radio app Apple podcast or wherever you get your podcast. Okay. So one of the questions I wanted to get answered at the top of this episode is like whether music that has AI in it can be quote unquote good or if the inclusion of such a process

47:38kind of disqualifies it for you guys. I mean, it's so much this is subjective question like people are going to feel differently about this. But I think we all landed in this similar place of like even if there was an AI song that I thought was good and catchy, I would still rather it have come from a real person if that makes sense. Like if we find out today that like some song I like was actually AI and I didn't know it, I would still wish that a human had made it. But it's still early days. Yeah. It may be a couple of years from now where most of the music we're listening to, if not is completely AI has so many AI elements that it becomes, you know, I think that's what Charlie is talking about protecting a bit. It's like it may become a bit of an alarm and it's like, okay, you don't want to listen AI music then you don't listen to music. Yeah. Period. Or a popular music. Yeah. I went into this, you know, when we were talking about this a few months ago during this episode, I was like, I will never listen to an AI song. I listened to real musicians in real art.

48:41And then like I said with this 10 10 bench thing is like it kind of is like, oh, it feels, I still listen to the song. I still like the song. And you know, that's not so cut and dry. I think there will be artists who will apply AI in a way that is not kind of quote lazy. Yeah. That will be hard to discern. And it may be the thing that like maybe it wasn't their original idea. Maybe, you know, like Charlie said, a lot of artists are using this to demo things. And maybe the AI does a cool thing with the guitar that a person ends up playing later. So I think the AI is going to be part of the creative process. I think it's inevitable. I think our tolerance for it is going to be tested in real ways. And I think if it is going to be in some ways inevitable, we have to hold high standards for musician to are going to use it. Like why wouldn't a person just do this thing? Exactly. Prove why AI was necessary in this process.

49:42Yeah. Or like music that is kind of signified as a human made. Yes. Like maybe in the future we'll have to classify it. That will become more valuable I think I guess. I largely agree where and what Charlie was saying where there's still needs to be it right now anyway, especially there has to be like this curatorial aspect to it where you can you can tell it to make something but someone has to decide if that's working or not. It's all those choices that actually usually create what we like. Like it's not at the point where if you have no musical ability, you're not going to be able to make a great song. Yeah. Like that's just it's impossible. Basically. And for a tool to make a like the Nashville example makes a lot of sense where I wrote the song. Here's a demo of it with a backing band through AI to kind of show what it might sound like. And then ideally then they get the resource to make the real thing. Yeah. That makes perfect sense is like this is not a real thing. This is just like a stopgap. Yeah. Or the mixing and mastering thing where it's like, okay, I'll do this as a first pass

50:44but then I might want to go and maybe the guitar needs to be louder here or that. And that's the real knowledge that you need. Yeah. And that'll get easier. I'm sure as these tools develop, but it's hard to imagine right now anywhere where it's going to be like type it in. I don't know anything about music. Yeah. Here's a great album. Yeah. Yeah. Exactly. I wonder too which genres this generative AI will be hitting first like pop music. Yeah. Hip hop. Like they're already so digital that the AI might not have as much of a problem recreating that. But in my little experiment yesterday with the No Such Thing theme, I did ask it to make a country version and it really didn't make any sense. Kind of like, I think what it needs to replicate kind of like really analog stuff. Yeah. It's going to struggle a little bit. Yeah. I would guess like mainstream pop and like electronic music just because it's already in that world of digital uncannyness. Part of it. Yeah.

51:44Yeah. So it's like that could be a positive. You know what it makes me think of is like we're going to have to have more sort of like the people are going to be good at using AI is going to be like the Beyonce's of the world, right? Where it's like Beyonce. You know people love to talk about she has so many songwriters probably. But at the end of the day, she's the person making the decisions, right? Yeah. There's a reason a lot of these artists who write for Beyonce and working these camps, their music does not do as well. Yeah. And it's like there is something that she is able to say, okay, I'm taking this element and putting together with this thing and putting together with that thing and then maybe that's going to be what people start to do with AI where it's like it's going to generate a bunch of stuff. You're going to have to have someone who has maybe they can't play that instrument or they don't have that idea for an lyric, but they're going to be able to when they see it pick it out and put it together into a new thing. Yeah. And maybe that initial thing wasn't human made, but the human touch is what makes it compelling in the end. No such thing as a production of kaleidoscope content.

52:51It is executive produced by Mangech Hattikader and it was created by Manny Fidel, Noah Friedman and Devon Joseph. The theme song, which does not use AI in any way, was created by me, Manny. Shout out to our guest, Charlie Harding, for joining us for this episode. He's the co-host of the Switched On Pop Podcast over at Vox. Definitely go check that out. It's a great show. And part two of this episode will be on your feeds this Friday. Go take a look. As always, check the show notes for a link to our newsletter and for the phone number that you can call to ask us a question that we can answer. See you next time. On Big Lives, we take a single cultural icon. People like Jane Fonda, George Michael, Little Richard. And we pull apart the story behind the image, discovering forgotten interviews that change

53:55exactly how we see these giants of our culture. We're here for the messy, the brilliant, the human version of our heroes. I'm Emmanuel Jochi. At Kai Wright. And this is Big Lives. Listen to Big Lives on the iHeart Radio app Apple Podcasts or wherever you get your podcasts. Is there an epidemic against literature? More specifically, Black Literature. I'm Jack Wiespies Thomas and on this season of the Black Lits podcast, we're outside asking why. They don't want a whole bunch of little elevators, Malcolm X's and Martin Luther King's on the route. And when you look at a lot of different cultures, feel like we need to know where you came from. And considering this is Black Reading Month, the only real question is, what are you reading next? Listen to Black Lits from the Black Effect Podcast Network on our iHeart Radio app Apple Podcasts or wherever you get your podcasts. What up, this Jay Valentine co-host of the R&B Money Podcasts alongside my brother, R&B Legend Tank. Together we've built a space for artists, executives and fans of R&B music to reminisce

54:56and tell the real stories behind the music. I have no money. I was hood famous. They think you're going to get everybody out. We lived in a project. And real love came out. They hear them music and we had nothing. I'm Smokey Robinson to Mary J. Blige, V2K to Jill Scott. If you're a super dedicated R&B fan, this is where you belong. Listen to R&B money from the Black Effect Podcast Network on the iHeart Radio app Apple Podcasts or wherever you get your podcasts. Hey, it's Bobby Beones. Join me and former NFL quarterback Matt Castle every Wednesday on our podcast, lots to say with me, Bobby Beones and Matt Castle. You're in training camp and the rookie quarterback has one good throwing session in front of the media. Suddenly everybody online says he should start over you week one. How do you handle this? You just go back out to practice the next day and wait for him to mess up. Listen to lots to say with Bobby Beones and Matt Castle on the iHeart Radio app Apple Podcasts or wherever you get your podcasts. This is an iHeart Podcast. Guaranteed Human.