8 minute read

Music and AI: A Conversation

Next Article
SUPPORT WKCR

SUPPORT WKCR

by Ale Díaz-Pizarro & Ted Schmiedeler

On April 4, 2023, TikTok user ghostwriter977 released the song “Heart on My Sleeve,” which featured artificial intelligence (AI)generated vocals of artists Drake and The Weeknd. The song was taken down from all streaming platforms by Universal Music Group, but not before garnering considerable attention on social media and sparked a debate surrounding the ethics of using AI in music. Weighing in, the following conversation between Station Manager Ale DíazPizarro and Student Life Director Ted Schmiedeler is adapted for publication from a News & Arts segment that aired on May 17, 2023.

Ted: There’s this AI generated Drake song called “Heart on My Sleeve” by AI Drake featuring AI The Weeknd. And it goes crazy. Like, it is a great song. The AI did a great job. Not only replicating a Drake song but creating a new, pretty good, pretty above average Drake song.

Ale: We don't even think it's legal to play. And the thing about it is, we're not even sure. I mean, it's probably not legal, but if you were to pursue legal action against us for playing this weird AI generated song, who was it even a copyright infringement against? The makers of the AI tool, or the fake AI Drake, or Drake himself—because it was, you know, samples of his music that were used to create that AI generated song? Who do you even pursue legal action against?

T: Because the voice on the song sounds exactly like Drake's voice. But it's technically not Drake's voice, it's just an almost perfect replica created by computers. It's like if I were to randomly have almost the identical voice of someone else that I know, you can't copyright my voice then even though it's super similar to somebody else's. But in this case, it's created by a computer to sound like someone else. So it's not like Drake is saying the words, the computer is creating something to mimic the Drake voice. So it gets into a really weird gray area. And then with the beat and the writing—I think, for this song, it was written by a human, but I know there are other songs that are AIgenerated, like the AI writes the lyrics, and then they have a different AI, do the production or make the beat and then they put it all together with the AI voice generator, so just gets into like a really weird gray area. So who is really responsible for making this art?

A: But since we are primarily a music radio station, I do kind of want to talk about AI in music, because I think when we think about AI art, it's mostly limited to the realm of the visual, like AI-generated images or photos or paintings. But I feel like we've had, in the past few weeks, a news cycle that has kind of come to, I guess, similar trends. On one hand, you have this fake Drake song. And then on the other hand, you have this Ed Sheeran trial for allegedly copying Marvin Gaye's “Let’s Get It On” for his “Thinking Out Loud”—which he won. I'm really only playing devil's advocate here, because I don't think that AI music is something feasible. But if we've been saying, like, “Oh, AI only works from things that already exist,” a defense that a lot of musicians bring up when they are brought to trial over songs that sound very similar is, “Well, there are only so many chords or chord arrangements that a musician can produce. It's not really my fault that they sound similar, it was not with the intent of copying.” Couldn't a similar argument be leveraged for AI music?

T: I think a similar argument could be made. But I think that the difference is in the production of that art. I mean, Ed Sheeran won his case. So let's assume he was proven to not have copied, so he came up with that chord progression. And he made that song entirely on his own, in his own space, and did not build on anything that Marvin Gaye had produced. But if it was an AI, producing a song, inherently, it has to rely on songs that already exist to come up with the new song. In order for it to run its production process, it must copy from things that it already knows.

A: You said build on as kind of like a throwaway phrase, but my other kind of line of thinking with this are samples. Aren't samples also a form of directly taking from something else?

T: I love this quote on samples. It's from RZA, de facto leader of Wu Tang Clan: he says that his goal is to use “the sampler more like a painter's palette than a Xerox.” Don't just directly copy it to make the song better, but make a sample like a drum or make it like a chord progression or something like that. Make it like a piano, make that an instrument in and of itself.

A: Well, I don't know if this counts. I mean, I guess it's a sample, but it's entirely composed of samples. Have you listened to the Grey Album? Danger Mouse? And it's just Jay-Z samples from the Black Album and the Beatles’ White Album? Which is why you can't get a record copy of it anywhere. But that one might be a good one to discuss, because you could argue that there's no original material in it in the sense that it's all Beatles samples. It's all Jay-Z samples. And yet, what's unique about it or original about it is the method of production. So I guess that gets us back to the same, you know, “it's process, not product” kind of cliché sentiment there.

T: But I do think it's valid. And one thing is, I feel like sampling is so cool. It's such a creative method of flipping something that's already produced into something that's newly produced. But one thing I think that's super important with samples is you have to get the sample cleared with the original artist. And that's one thing that thus far the AI is not doing. This random guy on the internet is not like clearing this with Drake at all. And Drake definitely is not giving permission. So it still feels kind of like the Wild West out there. That’s the thing with AI, I think it can be used for a lot of nefarious things, for example easily cheating on assignments. My mom is a high school English teacher, and when, the first weekend after the first big wave of news articles came out, she ran a bunch of her students’ assignments through an AI checker or something like that, and she caught a few kids doing it.

A: So your mom has been able to detect some of these AI-generated essays. They're kind of transparent, some of them, and that's what's scary about the Drake song, is the fact that it's not quite so transparent.

T: Yeah, it's definitely not transparent.

A: I don't really know what to make of AI music. Because I feel like that is such a different realm than for visual art and, and so on. We've talked about samples, remixes. Even just songs that sound like one another,— feel like music is a is a medium that's uniquely conducive to reinterpretation and rewriting. In jazz, as in country music, there are standards, and it's all about your own take on the standard. Why couldn't there be an AI take on this? I don't know. Again, I'm not subscribing to this, but I am concerned about the kind of arguments that might be leveraged in favor of AI in music or something similar.

T: I think it's so hard because it's not like above ground yet. The only places that it exists are

TikTok and YouTube where anyone can really post anything.

A: Well, but you could have someone saying, like, “Well, because AI generated music is not created by an artist, you're not losing out on any royalties but you are benefiting—like all the people who are listening to it and enjoying it.” In that scenario, I think opposing AI becomes more of a value thing or an existential thing over benefits.

T: It's super bleak. If in, like, 20 years, everyone is just in their headphones, and it's just a different AI making their music, you totally lose an aspect of music in the production process of a human putting effort into making this.

A: Well, I do think that when algorithms and AI DJs are proliferating, there's real value in places like radio stations. Which is why I don't believe that radio is dead or dying. That's why a lot of the “music industry,” quote-unquote, is kind of in a tailspin right now, because you're not really making songs anymore to get on the Top 100. You're making songs to go viral on TikTok instead of climbing the charts. That just means pop looks different. But to me that's not necessarily an indictment of where music is headed as a whole. I think there've been some incredibly ambitious and worthwhile musical projects in the last couple of years, in the last year, even, that have had absolutely nothing to do with climbing the Top 40. But if it becomes the fact that streaming and pop increasingly turn to algorithms to recommend content—and I hate the term content to refer to creations, but some of it is just content—then that just highlights the value of curation even more.

T: Yeah, I totally agree. I very much echo the point about how there is still music out there that is not just for the algorithm. You know, I think one common critique is the, you know, the songs are getting shorter, the albums are getting longer. I think it's about what the goal of the artist is, like, if you if you're just trying to run up your numbers, you know, you're gonna make like a 22-track, hour-and-10-minute album, and you're gonna hope that there's two or three million-plus songs in there. But if you're trying to make music that speaks to you, speaks to an audience, then you're you're more willing to do the 40-minute album, or the hourlong album with only ten songs on it, or something along those lines. Yeah, I think the music it's still out there, even if it's getting harder to find. And again, that gets back to the point about curation and the importance of radio and all that.

A: I definitely agree. And I also think, you know, places that still kind of tie the new to the old in the way that KCR does—and anyone reading this does not need us to preach to the choir and talk about how great KCR is, hopefully. But I've had the chance to put together such great shows that have had thought behind them, and meaning, and have allowed me to learn something from them. And I think it's it's so different to come at something with the goal of discovering or helping others discover something, rather than just pushing stuff or selling you on something. The fact that the way to get your song to be big now is to make sure that it gets on some Spotify playlists somewhere. I do think art has nothing to do with sales numbers. And again, that's very romantic, what have you, but—

T: I don't know if that's romantic. I will fight tooth and nail that sales do not equal quality. I will die on that hill. My brother loves to use the argument of, "Oh, well, you know, this person has more streams than this person per month." I'm like, that is a horrible argument. It's terrible.

A: It is. I mean, I don't love to talk about AI, but I'm glad that I don't feel like a Luddite anymore. And of course, anyone still listening to analog radio? Probably not going to vehemently disagree with us. But it's good to know that, as AI art gets pushed as this inevitable, revolutionary thing, KCR is still a home for music beyond its increasing commodification.

This article is from: