Loading summary
Transcript1 lines
- [00:00]
A
Welcome to the podcast. Today we're talking about deezer that says AI generated tracks are now more than 50% of all their daily uploads. So half of all music going up is AI. And for those that don't know Deezer, it's essentially a Spotify or Apple music competitor. It's the biggest one in France. When I, you know, I lived in France for a couple years, that's basically what everyone listens to music on. And a lot of different places in Europe, it's pretty popular. And the reason why I think it's important is because I think it's indicative of what's happening in a lot of different geographies, not just France, but you're probably seeing something very similar on Apple and Spotify. So we're getting into the state of all of all of this with AI and music generation. But before we do, I want to tell you about the AI Hustle School community. Every single week, Jamie and I, we go and record a bonus piece of content we put on the AI Hustle School community, and we actually record a live stream right now. For the end of the month, it is free to join the AI Hustle School community. It won't be free forever. And with the free membership, you get 5,5 of our bonus videos. You get live streams every single week for free, and you get access to a community of over 570 members. We have over a hundred videos we've recorded of how we grow and scale our businesses using AI tools. The numbers, the metrics, the breakdown, all the stuff going to share publicly, that's $19 a month if you want to get access to the library of over 100 videos. But otherwise, you can join the school community for free, and you can get access to tons of incredible members and amazing community. Okay, let's talk about what's going on with Deezer. So if you. This is a company that said that they're going to delist AI tracks that are unstreamed for six months because they said that they're getting 90,000 uploads per day. And so they're like, look, if you don't get any streams on your song for six months, we're just going to unlist you. Because we have so much music, just the volume of it is insane. And I'm going to be honest, it's, you know, using a company like Suno, it's very easy to create a song that sounds good. And it's interesting because I know people have a lot of different opinions on this. I personally make a lot of money every single month from Spotify. A chunk of my catalog I made before AI was without using AI essentially. And a chunk of my catalog I've used with AI. Some songs I've remade with AI since, you know, like a lot of stuff I made when I was in college. My, my. The biggest part of my catalog is I would just basically make relaxing piano music. I had a friend that was good at just playing relaxing piano music. I'd get him to record for a couple hours. I take that, chop it into two minute chunks and just post it as songs online to study to or relaxing music. Most famously. A lot of people laugh that I have a lot of music labeled relaxing music for your cat to help your cat relax. And I make $600 a month and I have since I was in college on like an album of relaxing piano music for cats, which is kind of funny. So what's interesting about all of this though is I do make a lot of money from AI, but it's also stuff that's getting streamed and listened to. And I think there's a couple different arguments. Like I would say my preference is always going to be a real human singing. You can just kind of the voices sound better, the instruments sound better, the lyrics sound better. Like everything really sounds better about a human usually. And even if it's kind of unique or interesting, like that's the sound that you like to hear. But that being said, I have listened to some AI music. I was recently, you know, scrolling through Instagram, tapping on some Instagram stories, saw someone's story, there was a song in the background. I was like, yeah, that's pretty catchy song. Looked it up and sure enough, once I got it, I realized it was AI generated because by the way, this person was doing something pretty sketchy. So I think they, they use the name of another famous artist as the artist name. So that's how they got streamed on their song. But their song was pretty good. I actually really liked it. It was very wholesome, but it was AI generated. But the other day I was actually on YouTube and I found a whole playlist of this, like folk music and the style I like. And it was AI generated and I clicked it and honestly sounded great. So I, I listened to it for probably like half an hour. My wife was laughing at me, but I was like, honestly, sounds good. So I can see it both ways. And sometimes it's less obvious. If it's instrumentals, it's not very obvious, right. Relaxing piano music that's AI generated sounds just like Normal piano music to me. But if there's like singers and different instruments, like guitar twang, I can usually hear tell the AI generated. That's also because I'm generating a lot of stuff so I get a good understanding of this. So there is money to be made and different streaming platforms are taking it different ways. We have a title that said if you use AI generated music, they're not going to pay you streaming royalties. So you straight up just won't get paid on their platform. But at the same time, you know, a lot of people are making content that people do. Like now where is the place for AI in or music in? AI and AI and music and like what's the relationship with Spotify came up with a bunch of terms recently that basically, I mean at the end of the day it's, if people like it, if it's high quality, then it's going to be streamed. My personal opinion is if you have content that sounds amazing, that is beautiful, using AI to help you with, I don't know, like, like for example, my wife is an artist and she releases all her own music. And I'm, I'm not going to lie, one of the biggest struggles we have and it's just because this isn't our ballpark. I'm a podcaster and she's a full time mom for the most part. So she's really good at writing songs. They sound, you know, amazing. She plays them live. But when it comes to recording, the recording process, the mixing, the mastering, all of that is basically the bane of our existence. We hate it. And I haven't gotten around to paying like a producer to go and make it all sound good or even to record it all because sometimes I'm like, shoot. The AC unit was on when we were recording and there's like a hum in the background now we got to put some compressors on it. And I don't know really what I'm doing if I'm being honest. So the problem is I wish there was a way to, if she could just record her music, how she does put it in some sort of AI platform and it goes and like fixes everything for it to make it sound good. But it's still her, it's still her plane, I would say, you know, right on. And that's honestly what a lot of plugins even on a lot of these music editing platforms do. Like they're going and kind of manipulating things even there's like flex pitch or there's you know, melodyne that's like auto Correcting your, your pitch a little bit. So like, there's all these things, things and ways that quote unquote, AI or computer software has been like used in people's music forever and somehow when it's like a kind of somewhat autonomous, people don't like it. But I think if you're a creative artist and you're using AI to help you with your flow, I mean, you're, you're just using software, that's fantastic. I would even go so far as to say if you use Suno and just one shot at a song and you love it, I mean, it would be pretty cool if you could go find an actual person to record it and sing it and maybe polish or tweak a couple of the words that might sound weird or cringy or off and, you know, have like a real person make it. I think that's amazing. Now, right now I'm saying this all from the standpoint of AI generated music. I can, when I listen to it, I can tell it's AI generated. But that won't be the case forever. These models are getting better and better. The twang of the guitar is going to start sounding more natural. It's going to be just like how the tone of Claude when it talks has gotten so much better since the very early days. So I think a lot of things are changing there. What's interesting is the share of AI generated music has climbed from about 10, and that was about 10,000 tracks a day on Deezer in January of last year. So just last year, 10% AI. And it is now forward and then it went to 44%, which is 75,000 tracks a day in April and now we're at over 50%. So Deezer is going to remove all of the tracks. I mean, I basically think this is a server cost saving thing. There's so much content. If it's not getting streamed, we're just going to pull it because obviously it's not popular. I mean, I think that's even generous. I would say, you know, if a song gets only like 100 listens a year or less than 100 listens a year, I probably pull it too if I were them. But I guess they're trying to leave something in out there for creativity. There is some technology that flags outputs from Suno and Yu Dio, which are the two biggest AI music startups. Both of them have major copyright suits. Basically, Suno went and scraped like all music, and that's what they used to train their model, which by the way, I've been saying for years I knew they did this because their models were just, it just felt like it was, it was too good. If you've ever tried Suno, it's incredible. You can just make absolutely amazing music and you can one shot it. And so yeah, they, they went and scraped everyone's stuff. They also apparently scraped a hundred thousand podcasts and it's for this reason that they are now actually ditching the old models that they use to make that they used all that content to make and they're now retraining new models with content that they have licensed and that's what they're going to, that's gonna be their new models that they were put out. So you're only gonna be able to use them with licensed content models, which is the right way to do it. It's kind of interesting though. And I mean, I don't. Yeah, I mean obviously they did it the wrong way and quote unquote illegal way, but at the end of the day they probably couldn't have grown their company any other way. And I know that's really controversial to say, but even with Claude, like the only reason Claude stayed relevant for so long is because it had a better tone than ChatGPT. And the reason they had a better tone is because they went and pirated every book in existence. So they were just better at writing and better at tone. And that's what Claude's competitive advantage is. Is there other ways to do it? Should people do it correctly? Yes, yes, yes, obviously. But at the end of the day, could Suno have been a company? Let's say they were, you know, they were like they were unable to raise money because their model just sounded bad. So they stole everyone's stuff, made a model that sounded good, gonna have to pay millions of dollars of fines, but, but they were also able to make a model that sounded good, prove that the use case work. From the proof of the use case, they were able to raise all the money to go and actually license the content and go pay the fines. So I don't know, like it sounds kind of crazy and my only other argument on this would be, so I'm not saying that they should have done what they did or anything, but I mean, I don' any other way their company could have really existed realistically. The other thing that I will say and another argument on all of this is that the Chinese models do not have any of these kind of copyright restraints and they could go make, they could clone like Claude clones, they could download all the music like Suno downloaded, there would be no negative repercussions for them. So if, you know, some of these models were, didn't do that, we'd probably have just a bunch of really good Chinese models. Or maybe OpenAI would just be significantly less good with less competition and they'd kind of have tried to fill in some of those gaps in the market. I'm not, I'm not sure. It's definitely a very interesting area in the market. I will say that Deezer opened its detection tools to other platforms. They have AI detected music tools. They've let other people use that earlier this year. And they also released a scanner for Apple Music and Spotify playlists last month. So you can scan, not just so you can scan playlists that you listen to, to, to, you know, discover if there's AI generated music inside of them. So at the end of the day, I think disclosure is the key. If you're creating music and it's AI generated, it's gotta be disclosed. People should know. Beyond that, you should have licenses to your stuff. Which again, another interesting point is that Suno, right when they downloaded everyone's stuff, they had in their terms, like, hey, this is totally licensed, you can use this for whatever you want, commercial purposes. And I mean, to be fair, like I have done that. I've posted music that I generated on Suno online and I make money from it. And I mean, they had commercial purposes on my licensing, so they're the ones getting the lawsuits and they're the ones that have to fight it. But their users are fine because I mean, what, what did I know back then, right? So in any case, it's a very crazy world. It almost makes me wonder if I should go and like regenerate some of my songs once their new model that they have licensing comes out or if it matters or what the difference is. But anyways, crazy world with all of this AI generated stuff. If you're interested in making money from AI generated music. I have an entire course in the school community, so there's a link in the description to our school community. It's $19 a month if you access that course and over a hundred other videos. I make about $3,000 a month right now just passively from my music catalog and there's a lot of ways to do that and a majority of it could be done today with AI. So if you want to check that out, the link is in the description. Thanks so much for tuning in guys, and I'll catch you all in the next episode.