Are we moving from the "AI can never match the quality of human artistic output" phase to the "oh, it's no big deal, art like that could always be faked" phase?
Tl;dr - in my opinion, a world where machines make our art is a smaller and sadder world.
I make music as a hobby so I’ve spent a lot of time thinking about this. My hot take is that a lot of music has been manufactured for years. My problem with AI music or art in general is not that it’s manufactured, it’s that they’re going to do their best to make it so that art isn’t something people make because an AI songwriting machine will be cheaper and never get tired or burnt out or messy the way a person will.
What has happened for years with manufactured acts is that someone finishes some music. Then you give it to a top liner or a writer’s room who play madlibs with whatever they’re writing about until the prosody feels right. Then you give it to a talented and entertaining person to perform. So you can sneak in new techniques, ideas, thoughts, feelings, performance nuances, whatever and legitimately move art forward. Not every song will and that’s ok. Not every artist and that’s ok.
When you have a machine doing it, first you’re going to get blatant propaganda like this often and they’re going to put it in genres that are usually subversive (punk, they’re coming for you again like they did when Nazis tried making punk). Second, you get nothing new. Machines know what has been, not what will be. They don’t yearn to add emphasis on just the right spot to make a song touch you in your soul. They don’t see that clever phrase in an otherwise unremarkable song and say “yeah, were making that the centerpiece”. LLMs will resolve a story. Suno and similar will use the same instruments the same way every time. They’ll cold open songs with just the barest difference in notes from the last song. They’ll make the exact production choices Rick Rubin or Butch Vig or Max Martin or Pharrell Williams make instead of taking inspiration and learning taste.
AI music, once it hits its stride, won’t be terrible. It’ll just be flat. And it’ll be everywhere. Acceptable at scale will crowd out excellence at the margins. And the only people making human music anymore will be people with enough money not to care that they don’t get paid or underground artists that people never get to hear because no one will elevate the subversive until it’s gotten too big to ignore. Rap had to elevate themselves. Punk, same. “Grunge”, as much as I hate that for a label for the conglomeration of everything from alt metal to post punk to updated standard rock just because they all wore flannel, same.
It’s up to the listener to accept or reject it. A lot of folks don’t care where something comes from as long as they enjoy it. And that gives people making music a hard row to hoe. There’s already enough hurdles to making music (and not just deep, thoughtful stuff, but people doing dumb, fun shit because it’s dumb and fun) that others will hear. Adding the Nashville, New York, or LA machine behind AI music just makes it harder for someone to feel like what they’re doing is worthwhile, which means fewer people do it.
They should be working on AI football teams. At least then there would be less CTE and gladiator fights in the world.
The video is clearly AI slop but worse, the track has all of the tells of an AI generated garbage, including the noise, artifacts, qacky compression, etc. Looks like slop. Sounds like slop.
And this shit proves how talent-less they are. See, a liberal with an agenda would just finance an artist that aligns with their views, but the GOP has to make artists up wholesale because their are no good MAGA artists to finance.
I get it. AI music bad, but let’s step away from the black and white thinking for a second, take off our Lemmy hats and think about a more nuanced approach.
Where do we draw the line with sampling? Where do we draw the line with borrowing from other people’s music? There is no legal protection of chord progressions. Artists are sampling whole songs, they’re sampling musical phrases, melodies and patterns, even recordings of individual notes of instruments.
We don’t know to what extent AI was leveraged to generate the music and lyrics of this song (maybe they wrote lyrics and fed it to a text to speech model, maybe they fed it specific audio to be used), but what we do know is that millions of artists everywhere borrow from and even directly use creative content from other artists.
Obviously, most Gen AI is trained on stolen content, but I’m offering the opportunity to see the potential of AI and music done in an (currently imaginary) ethical way.
Again, where do we draw the line at sampling in music?
I wasn’t wearing my “Lemmy hat”, I was wearing my music production hat. There are a lot of generative music tools out there that do a fairly decent job when part of a mix of tools. You can use some pitch correction and re-voicing tools to keep the vocalist on pitch and change their voice entirely. Instead, I can almost guarantee that they wrote the lyrics and fed them to Suno along with a genre prompt and saved personality. I doubt they bothered to even use the Suno Studio feature. This genuinely sounds like default generative output. It’s lazy, horribly mixed, noisy, and all around minimum effort.
PS: I also have not doubt that the lyrics are also prompt generated.
Music production hat makes you equally resistant to AI usage. With my music production hat on it’s equally difficult to listen to. Super compressed. That noise is ridiculous. The song itself pretty basic. BUT it’s not bad. I’ve heard human make stuff that sounds wayyyyy worse. Obviously it’s stealing from a lot to create something halfway decent. It’s only gonna keep sounding better (think autotune vs melodyne) and/or our ears are going to adjust to this quality of music. As I’m sure you know, music is heavily compressed/limited and way louder than generations past. And loooootts of overproduced artists in the radio. Part of me feels the same way. I’ll never like AI generated music. Then another part of me say, what if it sounds good? Why not? I still listen to MJ sometimes and he diddled kids.
BigMacHole@sopuli.xyz
on 14 Mar 20:23
nextcollapse
As a Free Thinking Conservative who Does Their Own Research TM this is FALSE! He’s OBVIOUSLY Real and I’ve SEEN him with my OWN EYES!*
*Someone I know said their BROTHER"S FRIEND"S UNCLE saw Him! RESEARCH!
threaded - newest
I mean, aside from the malignant racism, just how different is this from Milli Vanilli or the Monkees?
TIL Milli Vanilli and the Monkees were not actually real
Real… but manufactured and mass marketed. Not organically grown.
Pretty sure they were carbon-based
The Monkees were actual musicians that did their own music.
People said they monkeyed around.
FWIW I included them precisely because I like their music (and their take on Neil Diamond’s music, and their cool AF trippy movie)
Are we moving from the "AI can never match the quality of human artistic output" phase to the "oh, it's no big deal, art like that could always be faked" phase?
Tl;dr - in my opinion, a world where machines make our art is a smaller and sadder world.
I make music as a hobby so I’ve spent a lot of time thinking about this. My hot take is that a lot of music has been manufactured for years. My problem with AI music or art in general is not that it’s manufactured, it’s that they’re going to do their best to make it so that art isn’t something people make because an AI songwriting machine will be cheaper and never get tired or burnt out or messy the way a person will.
What has happened for years with manufactured acts is that someone finishes some music. Then you give it to a top liner or a writer’s room who play madlibs with whatever they’re writing about until the prosody feels right. Then you give it to a talented and entertaining person to perform. So you can sneak in new techniques, ideas, thoughts, feelings, performance nuances, whatever and legitimately move art forward. Not every song will and that’s ok. Not every artist and that’s ok.
When you have a machine doing it, first you’re going to get blatant propaganda like this often and they’re going to put it in genres that are usually subversive (punk, they’re coming for you again like they did when Nazis tried making punk). Second, you get nothing new. Machines know what has been, not what will be. They don’t yearn to add emphasis on just the right spot to make a song touch you in your soul. They don’t see that clever phrase in an otherwise unremarkable song and say “yeah, were making that the centerpiece”. LLMs will resolve a story. Suno and similar will use the same instruments the same way every time. They’ll cold open songs with just the barest difference in notes from the last song. They’ll make the exact production choices Rick Rubin or Butch Vig or Max Martin or Pharrell Williams make instead of taking inspiration and learning taste.
AI music, once it hits its stride, won’t be terrible. It’ll just be flat. And it’ll be everywhere. Acceptable at scale will crowd out excellence at the margins. And the only people making human music anymore will be people with enough money not to care that they don’t get paid or underground artists that people never get to hear because no one will elevate the subversive until it’s gotten too big to ignore. Rap had to elevate themselves. Punk, same. “Grunge”, as much as I hate that for a label for the conglomeration of everything from alt metal to post punk to updated standard rock just because they all wore flannel, same.
It’s up to the listener to accept or reject it. A lot of folks don’t care where something comes from as long as they enjoy it. And that gives people making music a hard row to hoe. There’s already enough hurdles to making music (and not just deep, thoughtful stuff, but people doing dumb, fun shit because it’s dumb and fun) that others will hear. Adding the Nashville, New York, or LA machine behind AI music just makes it harder for someone to feel like what they’re doing is worthwhile, which means fewer people do it.
They should be working on AI football teams. At least then there would be less CTE and gladiator fights in the world.
The video is clearly AI slop but worse, the track has all of the tells of an AI generated garbage, including the noise, artifacts, qacky compression, etc. Looks like slop. Sounds like slop.
No one ever accused MAGA of being intelligent.
Most MAGA have some sort of brain damage that causes them to not detect AI slop.
Lead poisoning, mostly
Lots of inbreeding too.
And this shit proves how talent-less they are. See, a liberal with an agenda would just finance an artist that aligns with their views, but the GOP has to make artists up wholesale because their are no good MAGA artists to finance.
I get it. AI music bad, but let’s step away from the black and white thinking for a second, take off our Lemmy hats and think about a more nuanced approach.
Where do we draw the line with sampling? Where do we draw the line with borrowing from other people’s music? There is no legal protection of chord progressions. Artists are sampling whole songs, they’re sampling musical phrases, melodies and patterns, even recordings of individual notes of instruments.
We don’t know to what extent AI was leveraged to generate the music and lyrics of this song (maybe they wrote lyrics and fed it to a text to speech model, maybe they fed it specific audio to be used), but what we do know is that millions of artists everywhere borrow from and even directly use creative content from other artists.
Obviously, most Gen AI is trained on stolen content, but I’m offering the opportunity to see the potential of AI and music done in an (currently imaginary) ethical way.
Again, where do we draw the line at sampling in music?
Samples have to be cleared by the original author to be legally used in a new track
Yeah lol this isn’t new news.
I guess you didn’t read the whole comment. It’s okay. It’s long.
Yeah that’s why I specified “in an ethical way”. So if gen AI tools paid royalties to artists you’d be cool with it?
I wasn’t wearing my “Lemmy hat”, I was wearing my music production hat. There are a lot of generative music tools out there that do a fairly decent job when part of a mix of tools. You can use some pitch correction and re-voicing tools to keep the vocalist on pitch and change their voice entirely. Instead, I can almost guarantee that they wrote the lyrics and fed them to Suno along with a genre prompt and saved personality. I doubt they bothered to even use the Suno Studio feature. This genuinely sounds like default generative output. It’s lazy, horribly mixed, noisy, and all around minimum effort.
PS: I also have not doubt that the lyrics are also prompt generated.
The biggest tell is the exact 2:00 length haha.
Music production hat makes you equally resistant to AI usage. With my music production hat on it’s equally difficult to listen to. Super compressed. That noise is ridiculous. The song itself pretty basic. BUT it’s not bad. I’ve heard human make stuff that sounds wayyyyy worse. Obviously it’s stealing from a lot to create something halfway decent. It’s only gonna keep sounding better (think autotune vs melodyne) and/or our ears are going to adjust to this quality of music. As I’m sure you know, music is heavily compressed/limited and way louder than generations past. And loooootts of overproduced artists in the radio. Part of me feels the same way. I’ll never like AI generated music. Then another part of me say, what if it sounds good? Why not? I still listen to MJ sometimes and he diddled kids.
As a Free Thinking Conservative who Does Their Own Research TM this is FALSE! He’s OBVIOUSLY Real and I’ve SEEN him with my OWN EYES!*
*Someone I know said their BROTHER"S FRIEND"S UNCLE saw Him! RESEARCH!
MY FRIEND’S HALF BROTHER’S UNCLE’S MOTHER’S BROTHER works at NINTENDO and told ME he was REAL!!!
This might very well be 40% loss money laundering.
Hold on to something, it’s gonna be more and more common.
Honestly, it would be wise for us all to just go analog for a while.
Real audience or botted audience?