I’m compelled by the idea that, like in Liz Pelly’s book, where one of Substack’s business goals is to reduce COGS, this could easily be a test to see if, like the nameless/faceless commodotized chill wave music they apparently foist on so many listeners, this could be a COGS-less solution to incremental revenue. I mean, sure, you might want to listen to Thee Osees, or Pink Floyd, if you’re a human at home being a human, but what about dentist offices and CVS stores? I’m only sort of kidding
I mean, I think it can be any of these things. It is so craven and so record geek specific, I can’t help but think it’s an experiment of some sort, and not a ploy at faking yr way through pop music stardom or at least interesting revenue deposits
Labeling should be a must. But I also think an embedded cue should be considered where a track can be identified as AI when heard where a label can't be viewed, like when using Shazam to identify a track.
Would be helpful to identify audio or video deep fakes too
If a serious musician "invents" a band via AI to release music they legitimately wrote and recorded, is that fraud? Of just a high-tech manifestation of what the Monkees were originally conceived to be by cynical businessmen? What about an avant-garde performance artist trying to prove a point about how much easier it is for good-looking (fictional) people in stylish clothes to become a streaming phenomenon than folks who don't have that look? I can see a lot of room here for someone to out-score their average by co-opting the visual presentation, or subversive Andy Kaufman-types making an art out of seeing how far they can take this.
Fraud involves deception, using a lie to deceive someone so that they do something they wouldn’t otherwise do for financial gain or to harm another person. Passing off a computer generated track as a real band to make money from streaming and dupe the consumer is fraud. If a human wrote and performed the music but obscures their identity, that’s different I think. If the music is authentic but the persona is AI then that falls within a tradition of art.
That's why I never understood the Milli Vanilli fan backlash; the songs they liked didn't become crummy simply because it turned out a different face was singing them; what they liked about the music itself remained intact. (We can contend they were crummy to begin with, but that's beside the point). Light-years of technology later, it still boils down to the book vs. its cover.
A few thoughts: 1. I've heard much worse music from humans. 2. Would it be better somehow (or worse) if AI outperformed most singer-songwriters? Can we give AI a Grammy, will the artist be called an operator? programmer? prompt master? 3. Other musicians have ALWAYS analyzed song structures, keys, chord progressions, lyrics...on music they haven't paid for...radio, TV, CD mixes from your ex, and most recently free streaming. So now we have a machine to do all that work. Let robots do what robots do best. 4. The music industry IS AN INDUSTRY, built to make money if it can. If streaming services decide to serve shitty AI music (or shitty human-made music) to customers that don't care, I guess it's their business...literally. This would naturally give rise to NON-AI streaming services and the age of authenticity. 5. I listened to Velvet Sundown Eyes - Psychedelic Blues / Indie Rock and I didn't notice any tell-tale signs of AI. I encourage people to COVER their favorite AI creations and throw a digital wrench in the worx. 6. Artists don't NEED a label or streaming attention. They may WANT the money/exposure, but they don't need it to do the work. Record your songs and put them on the internet: DONE. 7. I thought about making this list 10 items long, but changed my mine. Trust me, this list was not created by AI...or was it?
Thank you for commenting on this and I agree, songs should be labeled as AI generated. I am also seeing AI created music that I would call parody music, where a common soul or country sounding track is produced with lyrics intended to be comedic or/and raunchy. It's like Redd Foxx wrote a 60s soul song. The songs use the software programs you cited. So let's say an AI soul song parody used Al Green, Sam Cooke, and other real soul acts (via machine learning) to create this new AI song, how is that any different than the sampling debate in the early days of rap?
I’m compelled by the idea that, like in Liz Pelly’s book, where one of Substack’s business goals is to reduce COGS, this could easily be a test to see if, like the nameless/faceless commodotized chill wave music they apparently foist on so many listeners, this could be a COGS-less solution to incremental revenue. I mean, sure, you might want to listen to Thee Osees, or Pink Floyd, if you’re a human at home being a human, but what about dentist offices and CVS stores? I’m only sort of kidding
Agree. What about the idea this is marketing?
I mean, I think it can be any of these things. It is so craven and so record geek specific, I can’t help but think it’s an experiment of some sort, and not a ploy at faking yr way through pop music stardom or at least interesting revenue deposits
Surprise - you’re cynical too!
Labeling should be a must. But I also think an embedded cue should be considered where a track can be identified as AI when heard where a label can't be viewed, like when using Shazam to identify a track.
Would be helpful to identify audio or video deep fakes too
If a serious musician "invents" a band via AI to release music they legitimately wrote and recorded, is that fraud? Of just a high-tech manifestation of what the Monkees were originally conceived to be by cynical businessmen? What about an avant-garde performance artist trying to prove a point about how much easier it is for good-looking (fictional) people in stylish clothes to become a streaming phenomenon than folks who don't have that look? I can see a lot of room here for someone to out-score their average by co-opting the visual presentation, or subversive Andy Kaufman-types making an art out of seeing how far they can take this.
Fraud involves deception, using a lie to deceive someone so that they do something they wouldn’t otherwise do for financial gain or to harm another person. Passing off a computer generated track as a real band to make money from streaming and dupe the consumer is fraud. If a human wrote and performed the music but obscures their identity, that’s different I think. If the music is authentic but the persona is AI then that falls within a tradition of art.
That's why I never understood the Milli Vanilli fan backlash; the songs they liked didn't become crummy simply because it turned out a different face was singing them; what they liked about the music itself remained intact. (We can contend they were crummy to begin with, but that's beside the point). Light-years of technology later, it still boils down to the book vs. its cover.
A few thoughts: 1. I've heard much worse music from humans. 2. Would it be better somehow (or worse) if AI outperformed most singer-songwriters? Can we give AI a Grammy, will the artist be called an operator? programmer? prompt master? 3. Other musicians have ALWAYS analyzed song structures, keys, chord progressions, lyrics...on music they haven't paid for...radio, TV, CD mixes from your ex, and most recently free streaming. So now we have a machine to do all that work. Let robots do what robots do best. 4. The music industry IS AN INDUSTRY, built to make money if it can. If streaming services decide to serve shitty AI music (or shitty human-made music) to customers that don't care, I guess it's their business...literally. This would naturally give rise to NON-AI streaming services and the age of authenticity. 5. I listened to Velvet Sundown Eyes - Psychedelic Blues / Indie Rock and I didn't notice any tell-tale signs of AI. I encourage people to COVER their favorite AI creations and throw a digital wrench in the worx. 6. Artists don't NEED a label or streaming attention. They may WANT the money/exposure, but they don't need it to do the work. Record your songs and put them on the internet: DONE. 7. I thought about making this list 10 items long, but changed my mine. Trust me, this list was not created by AI...or was it?
If Lana Del Ray and Mazzy Star had a baby, and that baby was stolen it would be named Velvet Sundown. - Abraham Lincoln
Thank you for commenting on this and I agree, songs should be labeled as AI generated. I am also seeing AI created music that I would call parody music, where a common soul or country sounding track is produced with lyrics intended to be comedic or/and raunchy. It's like Redd Foxx wrote a 60s soul song. The songs use the software programs you cited. So let's say an AI soul song parody used Al Green, Sam Cooke, and other real soul acts (via machine learning) to create this new AI song, how is that any different than the sampling debate in the early days of rap?