As someone who occasionally writes a bit of music, I've had trouble putting into words exactly my feeling around AI generated art. One of the best takes I've seen on it which perfectly captured my feelings was by the Oatmeal: https://theoatmeal.com/comics/ai_art
The Oatmeal gave it away in the first 10%. If you like it until you find out it's AI, then it's not about the art, it's about the viewer, their judgement, preference, and opinion, not something intrinsic to the piece.
That's fine, but every attempt to then justify that judgement, preference, or opinion, in a way that does reflect an intrinsic aspect of the piece is a lie, a post hoc rationalization that cannot withstand the simple fact that it was liked as a thing in itself before it's story of creation was known.
Let's say I introduce you to a friend of mine. You chat for a little while, and everything seems to be going well, you like them. Then, I tell you that my friend loves kicking puppies and infants. Really hard too, like he's kicking a field goal.
All of a sudden, you start acting like he's a "bad guy," who should "burn in hell." You liked him just fine before I gave you that info, so clearly that's just a post hoc rationalization that cannot withstand the simple fact that he was liked as a person before his interests were known.
What are you talking about? Knowing a person and knowing a painting are nothing alike. You're the first person I've ever met that has ever suggested similarity.
Person or painting is not relevant in this argument by analogy. It's the commonality of "more than meets the eye", or "don't judge a book by its cover", or "appearances can be deceiving" between the two situations. There's dozens of idioms for this situation because it's so common.
I could just as easily make a painting analogy. You go to an art gallery and a piece immediately catches your eye. The composition, colors, you even think you can see a rosy meaning behind it all.
Then the curator comes by and explains that it was painted by David Duke, grand wizard of the KKK, and that the piece represents his deeply racist hatred of black people and his desire to murder innocents.
What you feel after having that explained is not "post hoc rationalization" - you're just recontextualizing your previous feelings towards the piece now that you have new information. That is the argument.
You could explicitly declare a separate judgment of quality apart from whatever else that was context-dependent.
If an artwork references something outside of itself, like making social commentary, then you need to understand that context to judge the quality of that reference: like whether the commentary of the artwork is astute or not.
The quality you can judge independently of context basically has to do with esthetics and craftsmanship, complexity, detail, consistency with itself and the like.
And in real life that's not what I see. I see people making comprehensive judgements about a piece that change when it's revealed that AI created the peice. When people rewrite their opinions, it doesn't come across as honest or authentic.
This is because people are using the artwork as a proxy to praise or criticize the creator.
When they say that the work is intricate, showing great skill and perseverence (to pick some quality judgments), they mean that the artist has skill and perseverence. When that turns out to be an algorithm that is mashing up information from the training data, then people take that back; there is no artist there with skill and perseverence. The artwork itself doesn't contain skill or perseverence; it's just potential evidence that skill and perseverence were involved in its creation, which is not confirmed when it's an algorithm.
Nobody actually cares about the art; it's always about the person.
People idolize the arists, while pretending to be idolizing their works.
That's why an artist can commit "career suicide" with a neo-nazi rant, or possession of child porn, or whatever; it was never about the actual artistic artifacts.
Why was Bob Ross, the painter, regarded as kitch? In large part because he used tricks to create detail without deliberation; almost like an algorithm. You use this kind of brush, dip it lightly into white, dab it it a few times this way on the cavas and that way and, hey look, it's an ocean spray above the rocks: "anyone" can do it if they follow along. It's thew same thing: the dabbing thing and whatnot were like his AI; he just "prompted" the brush, make me a realistic spray and the brush did it with a trick.
Art has to be confirmed evidence of Hard and Deliberate Effort having taken place, using Skills that took Years of Practice combined with Talent to acquire; that's just the way it is.
Art fans are basically sports fans; art is a kind of sport, where any kind of shortcuts are like steroids, hidden motors and course cutting; those who retract their judgments believe that they have all of Honesty and Authenticity entirely on their side, because it's never about the Work by itself, but only circumstances of its creation.
>People idolize the arists, while pretending to be idolizing their works.
100% in agreement. No judgement for idolizing or at least appreciating hard work, talent, or skill. Just don't piss on my leg and tell me it's raining. Be honest about that reasoning otherwise it sounds like lies because that's what it is.
Suno music is the modern equivalent of sample packs that are like Lego blocks, to allow people to quickly throw together cliches.
It can pass for OK music and good musicians sometimes use stuff like that, too.
But it's much closer to these soulless "London Authentic Jungle" loop sample packs or to "Dance eJay" than to a musical instrument.
I also feel it's not very similar to generative electronic music, which is about control and experimentation in a very different way (generating cliches with generative synths is hard, apart from some ambient drones or meditative arpeggios).
Suno and similar tools are designed to generate cliches from... text.
It doesn't matter that it incorporates AI and does stuff that was thought impossible a couple of years ago.
It's still a pastiche of cliches and not a musical instrument.
And the comic is nice!
Others mentioned sampling and synthesizers here, and I can't think of a comparison that I'd consider more misplaced.
Good sampling is really creative.
And bad sampling is the same as a bland, commercial cover version of a song. Many people don't bother to differentiate, but sampling and cover versions are very different things, and it's rare for both to appear together in one good song.
I was replying to a post that has since been deleted, but I'll leave my reply here anyway, for posterity:
> Synthesizers are regular musical instruments.
That's what people think now, yes. But I remember when the attitude was: "Synthesizers are electronic and controlled by a computer — it's not really an instrument, there's no creativity to it." . The Uk Musicians Union even discussed banning them [1] [2]
I'm not saying Suno is exactly the same. Of course history doesn't repeat itself exactly, but it does tend to rhyme.
I'd note that AI tools have this quality where they look almost usable right out of the gate, which fools people. It might resemble something, but it's not necessarily your vision. The gap between "this sounds like music" and "this is what I heard in my head" is where actual craft lives.
Still, our track record for predicting where the "real music" line falls is pretty poor. Maybe some humility is warranted.
[1] https://www.muhistory.com/contact-us/1971-1980/ "A meeting of the Central London Branch in May 1982 passed a motion for an outright ban on synthesisers" (was never official policy ... but did get voted on)
I think this is not quite the right analogy. A better analogy is procedurally generated music, because that’s what model-generated music is. But just like with LLM code generation, the input to the program is natural language (or maybe multimodal image/audio/whatever), and the program is implicitly defined by learning from examples of music.
I think a lot of the issues are the same. Like you might expect the model to go off the rails if you venture away from the bulk of the training distribution. Or maybe the b most effective way to use it creatively is in some kind of interactive workflow revising specific chunks of the project instead of vibe-coding/composing from whole cloth.
I've tried using Suno studio for aiding my writing/production. I mainly play guitar and bass, and a pretty mediocre keyboard player. Drums I can play some, but programming drums is infinitely easier for me when I write music home. I also have a home studio in the basement where I can record all instruments (drums included).
My first motivation to try out Suno, was that since I'm programming drums and laying OK keys, maybe Suno will do better? And, yes, sure did. I still record everything else, and I'm the "driver" when it comes to the creative direction.
But, still, you only get so much freedom. Best approach has been to record some idea of a track, feed it to Suno, and hope that it'll come up with something close (but better) than what you can. I feel zero musical accomplishment from generating the tracks, but the end result sounds better than the tedious job of programming instruments yourself. So, not much different than pulling tracks from a sample pack.
From a one-man band perspective, it works well. But of course there are many concerns, both ethically and otherwise.
With that said, I've listened to a lot of purely Suno made music generations, and it all sounds extremely polished and extremely average. Average in the sense that it sounds like the average of the music it has been trained on. My first prediction was that modern country would be the first victim of this, fact being that for the past 10 years modern country has sounded like mindless AI slop, but made by humans.
Polished and average is a good description. There are also no highs or lows across a track - it is very samey all the way through, where an actual creative person will build some kind of arc into a song.
But it actually is music creation because I don't just write a prompt and accept whatever comes out. I listen to it, and if I don't like it, I press enter again! That's active participation!
As someone who occasionally writes a bit of music, I've had trouble putting into words exactly my feeling around AI generated art. One of the best takes I've seen on it which perfectly captured my feelings was by the Oatmeal: https://theoatmeal.com/comics/ai_art
The Oatmeal gave it away in the first 10%. If you like it until you find out it's AI, then it's not about the art, it's about the viewer, their judgement, preference, and opinion, not something intrinsic to the piece.
That's fine, but every attempt to then justify that judgement, preference, or opinion, in a way that does reflect an intrinsic aspect of the piece is a lie, a post hoc rationalization that cannot withstand the simple fact that it was liked as a thing in itself before it's story of creation was known.
Let's say I introduce you to a friend of mine. You chat for a little while, and everything seems to be going well, you like them. Then, I tell you that my friend loves kicking puppies and infants. Really hard too, like he's kicking a field goal.
All of a sudden, you start acting like he's a "bad guy," who should "burn in hell." You liked him just fine before I gave you that info, so clearly that's just a post hoc rationalization that cannot withstand the simple fact that he was liked as a person before his interests were known.
What are you talking about? Knowing a person and knowing a painting are nothing alike. You're the first person I've ever met that has ever suggested similarity.
Person or painting is not relevant in this argument by analogy. It's the commonality of "more than meets the eye", or "don't judge a book by its cover", or "appearances can be deceiving" between the two situations. There's dozens of idioms for this situation because it's so common.
I could just as easily make a painting analogy. You go to an art gallery and a piece immediately catches your eye. The composition, colors, you even think you can see a rosy meaning behind it all.
Then the curator comes by and explains that it was painted by David Duke, grand wizard of the KKK, and that the piece represents his deeply racist hatred of black people and his desire to murder innocents.
What you feel after having that explained is not "post hoc rationalization" - you're just recontextualizing your previous feelings towards the piece now that you have new information. That is the argument.
I understand the argument. It's just a bad argument.
If your judgement of a work of art changes with additional context, then your judgement of the art was not based in the quality of the art in itself.
You could explicitly declare a separate judgment of quality apart from whatever else that was context-dependent.
If an artwork references something outside of itself, like making social commentary, then you need to understand that context to judge the quality of that reference: like whether the commentary of the artwork is astute or not.
The quality you can judge independently of context basically has to do with esthetics and craftsmanship, complexity, detail, consistency with itself and the like.
And in real life that's not what I see. I see people making comprehensive judgements about a piece that change when it's revealed that AI created the peice. When people rewrite their opinions, it doesn't come across as honest or authentic.
This is because people are using the artwork as a proxy to praise or criticize the creator.
When they say that the work is intricate, showing great skill and perseverence (to pick some quality judgments), they mean that the artist has skill and perseverence. When that turns out to be an algorithm that is mashing up information from the training data, then people take that back; there is no artist there with skill and perseverence. The artwork itself doesn't contain skill or perseverence; it's just potential evidence that skill and perseverence were involved in its creation, which is not confirmed when it's an algorithm.
Nobody actually cares about the art; it's always about the person.
People idolize the arists, while pretending to be idolizing their works.
That's why an artist can commit "career suicide" with a neo-nazi rant, or possession of child porn, or whatever; it was never about the actual artistic artifacts.
Why was Bob Ross, the painter, regarded as kitch? In large part because he used tricks to create detail without deliberation; almost like an algorithm. You use this kind of brush, dip it lightly into white, dab it it a few times this way on the cavas and that way and, hey look, it's an ocean spray above the rocks: "anyone" can do it if they follow along. It's thew same thing: the dabbing thing and whatnot were like his AI; he just "prompted" the brush, make me a realistic spray and the brush did it with a trick.
Art has to be confirmed evidence of Hard and Deliberate Effort having taken place, using Skills that took Years of Practice combined with Talent to acquire; that's just the way it is.
Art fans are basically sports fans; art is a kind of sport, where any kind of shortcuts are like steroids, hidden motors and course cutting; those who retract their judgments believe that they have all of Honesty and Authenticity entirely on their side, because it's never about the Work by itself, but only circumstances of its creation.
>People idolize the arists, while pretending to be idolizing their works.
100% in agreement. No judgement for idolizing or at least appreciating hard work, talent, or skill. Just don't piss on my leg and tell me it's raining. Be honest about that reasoning otherwise it sounds like lies because that's what it is.
Soulless shit?
It is exactly this.
Suno music is the modern equivalent of sample packs that are like Lego blocks, to allow people to quickly throw together cliches.
It can pass for OK music and good musicians sometimes use stuff like that, too.
But it's much closer to these soulless "London Authentic Jungle" loop sample packs or to "Dance eJay" than to a musical instrument.
I also feel it's not very similar to generative electronic music, which is about control and experimentation in a very different way (generating cliches with generative synths is hard, apart from some ambient drones or meditative arpeggios).
Suno and similar tools are designed to generate cliches from... text.
It doesn't matter that it incorporates AI and does stuff that was thought impossible a couple of years ago.
It's still a pastiche of cliches and not a musical instrument.
And the comic is nice!
Others mentioned sampling and synthesizers here, and I can't think of a comparison that I'd consider more misplaced.
Good sampling is really creative.
And bad sampling is the same as a bland, commercial cover version of a song. Many people don't bother to differentiate, but sampling and cover versions are very different things, and it's rare for both to appear together in one good song.
Please "just type an AI prompt", then show us what music you've made with AI. I'll wait.
Is this like synthesizers or sampling all over again? (to wit: the whole "it's not real" song and dance)
I was replying to a post that has since been deleted, but I'll leave my reply here anyway, for posterity:
> Synthesizers are regular musical instruments.
That's what people think now, yes. But I remember when the attitude was: "Synthesizers are electronic and controlled by a computer — it's not really an instrument, there's no creativity to it." . The Uk Musicians Union even discussed banning them [1] [2]
I'm not saying Suno is exactly the same. Of course history doesn't repeat itself exactly, but it does tend to rhyme.
I'd note that AI tools have this quality where they look almost usable right out of the gate, which fools people. It might resemble something, but it's not necessarily your vision. The gap between "this sounds like music" and "this is what I heard in my head" is where actual craft lives.
Still, our track record for predicting where the "real music" line falls is pretty poor. Maybe some humility is warranted.
[1] https://www.muhistory.com/contact-us/1971-1980/ "A meeting of the Central London Branch in May 1982 passed a motion for an outright ban on synthesisers" (was never official policy ... but did get voted on)
[2] https://www.musicradar.com/news/the-union-passed-a-motion-to... Terrible writeup, but lists sources at least.
I think this is not quite the right analogy. A better analogy is procedurally generated music, because that’s what model-generated music is. But just like with LLM code generation, the input to the program is natural language (or maybe multimodal image/audio/whatever), and the program is implicitly defined by learning from examples of music.
I think a lot of the issues are the same. Like you might expect the model to go off the rails if you venture away from the bulk of the training distribution. Or maybe the b most effective way to use it creatively is in some kind of interactive workflow revising specific chunks of the project instead of vibe-coding/composing from whole cloth.
Where GenAI goes off the rails it starts being interesting for art, IMO :)
edit:typo
At least with samples, you had to do some actually arrangement - namely placing out the samples in your DAW, make some active creative decisions, etc.
With tools like Suno, you really don't need to do anything other than just type in the your prompt, and hope for the best.
The closest analogy I can come up with, is that making music with tools like Suno is like hiring a musician / composer to make the music for you.
I've tried using Suno studio for aiding my writing/production. I mainly play guitar and bass, and a pretty mediocre keyboard player. Drums I can play some, but programming drums is infinitely easier for me when I write music home. I also have a home studio in the basement where I can record all instruments (drums included).
My first motivation to try out Suno, was that since I'm programming drums and laying OK keys, maybe Suno will do better? And, yes, sure did. I still record everything else, and I'm the "driver" when it comes to the creative direction.
But, still, you only get so much freedom. Best approach has been to record some idea of a track, feed it to Suno, and hope that it'll come up with something close (but better) than what you can. I feel zero musical accomplishment from generating the tracks, but the end result sounds better than the tedious job of programming instruments yourself. So, not much different than pulling tracks from a sample pack.
From a one-man band perspective, it works well. But of course there are many concerns, both ethically and otherwise.
With that said, I've listened to a lot of purely Suno made music generations, and it all sounds extremely polished and extremely average. Average in the sense that it sounds like the average of the music it has been trained on. My first prediction was that modern country would be the first victim of this, fact being that for the past 10 years modern country has sounded like mindless AI slop, but made by humans.
Polished and average is a good description. There are also no highs or lows across a track - it is very samey all the way through, where an actual creative person will build some kind of arc into a song.
But it actually is music creation because I don't just write a prompt and accept whatever comes out. I listen to it, and if I don't like it, I press enter again! That's active participation!
/s