Yes, sampling is an artform. We are on the same page here.
But your original comment implies that using Suno would be like sampling.
Therefor I mentioned that you need to properly credit the usage of samples, what Suno is not doing, Suno is steeling from real artist.
Hope it is more clear now.
I had tons of these "visual styles" and I remember how MS liked to patch themeui.dll and uxtheme.dll often rendering theming broken.
There were also attempts at customizing XP booting screen to achieve the perfect "it's not Windows" effect but that could easily render installation unbootable
Meanwhile I've mixed a song of mine down on a Tascam 688 8-track tape recorder. I have a big smile on my face because I find this very enjoyable. The haptics, the sound and my hand-crafted product. A piece of art made by a human. No AI will replace this for me.
This argument always seemed weird to me because it's obviously the case that ‘intent and creativity’ is at best a slim minority of consumer preference. If I want a simple dubstep line with random romantic words meaninglessly peppered in it, or the same 4 chord pop song as every other pop song, or you know, anything popular, that's easy. If I want something that's saying something interesting in the topics I find interesting, or doing something new with the medium that's also well executed and listenable to on repeat, finding that is actually just hard, and it's rarely the stuff with the dollars and overlapping interests to get the high production values.
I'm not going to claim AI audio isn't also awash with popular themes and tropes, or that it's a bastion of creativity. I'm also not going to claim that the deepest, really creative ideas aren't expressed in human written works. There are enough people to make truly exceptional songs and prompt many truly mindless AI generations. And there's also nothing wrong with most songs optimizing for personal preferences that are not that; I'm not trying to 'argue against' popular music.
But I am going to claim, for me, that it just hasn't been practical to saturate my tastes from public media, and that most of the reason I personally listen to AI music is that I want something that says or does something I think is creative, exploratory, or intellectually interesting that I don't know how to get from anywhere else.
I was alive to hear the evolution of "dubstep" before it became a cliche (the wobble stuff). AI couldn't invent a new sound.
The way you describe music, sure, there will be an AI that is able to provide you with a continuous stream of audiotory stimuli, like the Penfield Mood Organ from "Do Androids Dream of Electric Sheep?".
That's just not what makes art or music interesting to me, and why I also don't listen to auto-curated "mood" playlists on Spotify.
> Penfield mood organ - Humans use the mood organ to dial specific emotions so they can experience emotions without actually possessing them. In the beginning of the novel, Rick implores his wife to use her dialing console to prevent a fight. He wants her to thoughtlessly dial emotions like "the desire to watch TV" or "awareness of the manifold possibilities open to [her] in the future" (Dick, 6). When emotions can be easily avoided with the mood organ, humans no longer require personal relationships to overcome feelings of isolation or loneliness.
I also don't want mood playlists, the majority of the time. I'm saying I use AI generations for exactly the opposite — so that I can explore and listen to things that are more intellectually interesting in the ways I find intellectually interesting, because the human music it's easy to find is common denominator music.
The way you describe AI ('continuous stream of auditory stimuli') is the way I'd describe Spotify. Sure, you could use AI to make a faux Spotify, but, like, why would you? The popular stuff already has saturating supply, and it will sound much better than an AI generation.
I was also having Spotify in mind when writing about algorithmic curation and GenAI in music ;)
Regarding this:
> I'm saying I use AI generations for exactly the opposite — so that I can explore and listen to things that are more intellectually interesting in the ways I find intellectually interesting
I just have not found any AI music that would satisfy this description. But I am very interested in failure modes of GenAI. Especially in Suno, it was cracking me up at times.
I'm also sure there will be a space for interesting and/or challenging music generated with neural networks involved.
But I don't see any revolution here so far.
Care to share examples of AI-assisted music you find interesting? To elaborate, I don't find jarring or curious combinations of cliches interesting.
AI could not invent a new style, it seems to me. To repeat this point.
And I've never had any problem finding interesting music.
Key to me is diving into labels, artists and their philosophy, after I got interested into particular ones (the other way around doesn't work for me).
I adore discogs.com for that. Regarding interviews and stuff, there's sadly a huge decline in quality written material about music, I feel.
"Lowest-common-denominator music" is exactly what Suno produces, at least in my ears.
I could go on and list music I like, but generally avoid that.
Wait, I'll do it anyway for a bit... at the moment, I like
Punctum - Remote Sensing EP
(Caterina Barbieri)
and
AtomTM vs Pete Namlook - Jet chamber LP
just for example
I also love so much other music.
To me, such music is miles apart from the slop I heard from AI.
I heard there's research into generating music in the style of JS Bach as well. How's that going?
I'd guess: probably bot too well, because the genius of Bach is not only in complexity, or counterpoint rules.
His music is very emotional to me (at least the portions I like).
And, like any good music, it has moments of surprise. It's not just a formula, or a "vibe", or a "genre".
Could AI create a new Techno, a new Blues, a new Bossa Nova?
I've been avoiding trying to class what counts as 'interesting' because it feels like the wrong point to make, for a medium so entwined with personal preferences that are so often at odds with each other. To single out my tastes feels like it would be saying 'paleontology is the true interesting subject, unlike architecture'. The tracks you listed here aren't really my thing, but I can see why AI audio doesn't help with it. That's fine.
I will also repeat that I'm well aware that the best stuff is definitely all human. It's not my genre either, but traditional composers like Bach certainly made extremely interesting, clever, even deeply-studiable pieces and AI 'in the style of' those composers surely won't capture much of that. There's a lot of stuff AI can't do wholesale; one particularly strong example is if you're Jacob Collier, AI is not going to make the complex harmonizations and song structures there.
AI is pretty bad at these textural or instrument exploration things like from Collier above or Mike Dawes or Yosi Horikawa or Yoko Kanno or Keiichi Okabe. There's a bunch of music I listen to because it's generically a genre or mood I like and it's well produced, which I won't list here, and AI audio can often do stuff like that at baseline but not especially well. There's also nostalgia; I'm also certain a huge part of the reason I like the Celeste soundtrack so much is in part that I liked the game so much.
But then there's a whole category of music I listen to where the texture is supplemental to the part that defines it, like most of Acapella Science or Bug Hunter or Tom Lehrer. Eg. Prisencolinensinainciusol isn't interesting to me because it's musically complex; the part I care about is that it's a listenable execution of an idea, not precisely how it was executed on. I don't keep coming back to I Will Derive by some random schoolkids recorded on a potato 17 years ago annually because it's sung well or they were particularly clever with how they took another song and changed the words; I come back to it because it's fun and reflects for me onto a part of my past that I remember fondly, and these things make me happy.
All these words and I've still only addressed half the comment. Ok, let's consider the idea that it's not enough for AI audio to facilitate the creation of interesting musical pieces, and it instead has to create whole interesting musical styles. I take issue with this in a bunch of places. I don't reject artists who I judge not likely able to create a new Bossa Nova. I judge artists based on whether the output they produce is something I want. I do the same for AI.
I also think the question about whether AI could 'create' a new style is somewhat misplaced. A style is a cultural centroid, not just a piece of audio. AI can definitely create new musical textures or motifs, but it's always being pulled towards the form of what it's being asked to produce. As long as we're talking about systems that work like today's systems, the question still needs to involve the people that are selecting for the outputs they want. Could that connected system create something as distinct and audibly novel as a new genre? Yeah, probably, given time and a chance for things to settle. That's a different question from whether it'll do so to an inspecific prompt thrown at it.
> I've been avoiding trying to class what counts as 'interesting' because it feels like the wrong point to make
Well, I kind of borrowed this wording from the description of your genAI experiences ("intellectually interesting").
I feel that it's nod a bad word to describe qualities of music, although it's a bit nondescript. Sure, music can be interesting, but still unpleasant etc.
But "interesting" means (to me) that the music makes you want to listen to it again.
Music doesn't need to sound pleasant. Or angry. Or sad. Or "abstract" (an oxymoron when it comes to describing a sound, still widely used).
Music is communication.
And just like it's a novelty and sometimes useful or entertaining to use ChatGPT, it's a novelty and sometimes interesting to use Suno.
That's pretty much it for me.
The magic of prompt => non-text media is also interesting, sure.
But not interesting anymore to me as art, at least not without being part of a bigger whole.
Good early example for this would be "Headache - The head hurts, but the heart knows the truth"
I frequently find myself hitting repeat. I have over a dozen I've listened to, according to the stats, more than 50 times each. By your definition that's interesting. QED?
I was alive to hear the evolution of "dubstep" before it became a cliche (the wobble stuff). AI couldn't invent a new sound.
The way you describe music, sure, there will be an AI that is able to provide you with a continuous stream of audiotory stimuli, like the Penfield Mood Organ from "Do Androids Dream of Electric Sheep?".
That's just not what makes art or music interesting to me, and why I also don't listen to auto-curated "mood" playlists on Spotify.
> Penfield mood organ - Humans use the mood organ to dial specific emotions so they can experience emotions without actually possessing them. In the beginning of the novel, Rick implores his wife to use her dialing console to prevent a fight. He wants her to thoughtlessly dial emotions like "the desire to watch TV" or "awareness of the manifold possibilities open to [her] in the future" (Dick, 6). When emotions can be easily avoided with the mood organ, humans no longer require personal relationships to overcome feelings of isolation or loneliness.
That being said, there is a spectrum, sure.
I am interested in generative music.
I do have scenarios where I listen to music and want it to blend into the background (e.g. soma FM).
But even then I love the short moment when a song comes up and I want to note it because it's distinct.
I am not interested in being robbed of that.
Also: why? Why, why, why?
Music is not just a recording packaged as a product. It is a thing humans do. And I say that as a person that enjoys mainly electronic music!
There are many talented humans, there is absolutely zero need for AI muzak, other than decreasing the price.
Musicians leveraging generative AI for creative purposes might become a thing and I am fine with that in principle, but the thought is a joke to me, as of now.
Creating audio from an idea is not the same as letting a machine create an interpolation of stolen ideas to match a prompt.
Culture is fluid. Music is about exploring the boundaries of what sounds good, often because of feelings. Related to the society in which the music is "consumed".
AI music is a commodity and generally uninteresting, like artists who only imitate styles.
But just like annoying over-commercialized music that only tries to scratch existing itches and match expectations, it can still work to a degree.
Intent is not a lofty concept, it's at the heart of what art is.
I was using ‘consumer preference’ here as a proxy for ‘the things that people tend to care about.’ The point I'm highlighting is that it's weird that when AI audio comes up, there's a reliable crowd of people discrediting it for missing a thing that clearly doesn't matter that much to most people in most contexts.
It's like, sure you can want things from music that are to your specific taste, but it's like coming into a post about, idk, a folk band and complaining that it's not metal. You're allowed to like your thing, but clearly most music is allowed not to be metal, why is this music specifically bad for not being metal?
And in this case the point I'm making is stronger, in that AI audio actually unlocks a lot of ability to listen things that are ‘interesting and creative’ but not widely available because of consumer preference, so it's actually more like showing up to a folk metal fusion band and saying the problem with this band is that it isn't metal.
You still totally miss the point of this thread. Making music and creating art for the sake of enjoyment. Consumer preferences are irrelevant here. I am totally fine if you like AI music. And I am sure someone can create art by interweaving AI elements. And keep in mind, the training data for Suno is coming/gets stolen from someone like me who creates real music. Personally I do not enjoy to make music with some proprietary 3rd party tools that can garden wall at every time, like Suno. But that is personal preference.
> Then tried to generate interesting music, failed spectacularly.
You can upload music and let suno arrange it in different styles. I'm a musician myself and am also interested in "interesting" music. I made experiments with my own music and was positively surprised by the "musicality" and creativity of the generated arrangements (https://rochus-keller.ch/?p=1350).
I could learn Danish with Anki, YouTube, ChatGPT and a teacher in evening school. I think Danish is even harder to learn then Persian. ;-)
Anki is a great concept!
How difficult a language is to learn is very subjective. My mother language is Danish and therefore learning German was relatively easy for me (relatively because it is always difficult to learn a new language!)
Manchmal, wenn ich ein Wort noch nie gesehen habe, kann ich die Bedeutung herausfinden, weil ich das entsprechende Wort auf Dänisch kenne.
Åh, du skriver en rigtig god tysk!
Du hast Recht mit der Ähnlichkeit der Wörter im Dänischen und Deutschen. Sind ja auch beides germanische Sprachen. Man muss aber aufpassen, es gibt einige "falsche Freunde". :-)
Z.B. Bier/bier, Kind/kind, Tag/tag, osv..
Har det godt!
If you know English danish/swedish/Norwegian is absolutely one of the easiest languages to learn.
People do underestimate how much a good teacher and putting in the work can make a difference. Money well invested if you actually NEED to use the language and will have opportunity to.
Having an example of too much of selectable text. When I copy a YouTube video title and paste it somewhere else suddenly the language code of the text appears in front of the pasted text line. That is also really annoying.
reply