Using AI to create music from other bands

Lounge
1

Sorry, the subject of this post was crafted to be a bit controversial. But I’m not trying to evoke anger or confrontation. :blossom: I’m just curious to see what people think. Let me explain…

I’ve been using Sonauto to generate music using prompts that mix together bands like Autechre and Ed Rush. It feels really similar to buying a Scratcher at the corner store. Most of the time you get something good. But sometimes you get something really good.

It’s undeniable when listening to this music that it’s coping a band’s style.

But it’s also arguable good music, evoking the same response that the band’s original 10 track albums evoked back in the day. It’s not accurate to call this a “cover”. Maybe “rip-off” might be more realistic. But even given that, if the music will bring pleasure to those who enjoyed that music from the past, is it acceptable?

It might be easy to take a firm stance on music from bands that are currently producing music. But let’s take someone like Front 242, from the Geography years. If we generate music like that because we love it and want to share more of it, how does that sit with people?

I’m curious because that’s exactly what I’ve been doing after 8:00 PM when I’m bored: I’m generating music either from bands in the past that I love, or creating hybrids using prompts that combine different styles. And sometimes the tracks sound fantastic. Would releasing them be disrespectful to the original artists?

I’d love to hear your thoughts.

1 Like
2

I’d say yes, because these models have been trained on their work without their consent and without them being compensated. If you just do it for your own personal enjoyment, sure why not (although we all know these systems use a lot of resources, which makes their use somewhat problematic).

I also find it hard to call these kind of generated works “music” perse. More a sophisticated kind of “muzak”.

8 Likes
3

Hi Bret

First of all, this has always been an interesting and “hot” subject, especially since the days where sampling and mashups became a technological reality.

The legality of it I want to put aside. I’m sure there’s lots of murky waters there and I’m sure that if you made a lot of money someone would find a way to sue you. But I don’t think that’s what you’re most interested in.

Then there’s using AI to make music. I’m not a fan but I respect that other people like to play around with it.

And then there’s the act of creating music, using various aspects of other peoples music. That’s a really murky area. The reponses from other musicians has historically spanned from “thanks for the flattery” to “no fck.n way, I’ll sue you, stop it”. I guess where I stand is like this: If you directly (or indirectly via AI) use material from other musicians, and you get their permission you’re good, otherwise you’re probably not good, if they find out. I think copying material and style are two different, but related things. Cover bands are a time-honored traditioin and they can be very good. Also, someone making something from scratch, in the style of someone else, is also quite Ok, and can be very good, although if you make lots of money someone will find a way to sue you.

I don’t think anyone cares what you do alone in your bedroom with your synths. If you start publishing it though, it’s another matter. And of course it also depends on how much/little of others’ material you have used.

If I were to finish with a simple formula I’d say: The more it’s your own creation the safer you are. The more it’s the creation of others the more dubious it becomes. I guess that’s my take on it.

And then my own philosophy: Anyone can create “a product” and call it music, and others might even like it. What is much more interesting is your own, unique creativity and how that translates into music. Put yourself into it, make it your own, create it from scratch if possible.

4 Likes
4

There is no shuffle… there is no shuffle here…

Sorry, got distracted…

Anyway… here are my thoughts:

A user prompts some model and it cobbles together, from existing data it has processed and out of noise, as output, something akin to what the user expects; the user enjoys; the user decides to release a track, or several, is it the user’s creation or just a convoluted way to plagiarize?

I will, like Lars, ignore the legal side for a bit; and I’ll discuss the human side first.

I think we are living times which are a bit sad: people just want to hear the same thing over and over again (I was waiting for my pizza the other day; 4 videos played, according to the lower thirds, 4 different artists… to my ears they all played the exact same song; only looked physically different… the videos were kinda similar as well :S ); but I still believe (hope?) this awful trend can be reversed. How? Releasing original music.

We all love certain bands and have a number of influences; learning to ape them is part of a musician’s journey; for example, the late Dimebag Darrell loved KISS, Ace Frehley, may he also rest in peace, in particular, he even had a tattoo of his spaceman face; he knew, I’m sure, quite a few of the Spaceman’s riffs; yet Pantera sounded quite different than KISS, even in their weird glam period; I love Frank Kleplacki’s work, specially the tracks written for real time strategy games, so I spent some time learning how to create something with a similar vibe.

Why not create your own and push original music to the fore again?

Fans of certain industries have been rather vocal in opposing AI use, even if it doesn’t comprise the whole work, a couple of examples: “Late Night With the Devil” is a movie I enjoyed very much; but it faced backlash because certain posters were AI generated (they’re not bad either… but people didn’t like how they were created); GZDoom, a port of Doom, finally got forked (after, apparently, years of putting up with a terrible attitude from the main maintainer… I have to disclose he was always nice to me) because he introduced AI code to the sources; several programs I use have been getting worse as of late: stuff that worked for years is suddenly broken; I have resisted the urge to write something akin to “Is this AI crap?” in my bug reports because… I wouldn’t like people writing that in reports sent to my projects (I don’t “vibe code”… not even a little)… so… there’s that.

Now… setting the human factor aside…

You mentioned Front 242; I suppose you are familiar with Razormaid’s work; AFAIK the reason it is rather pervasive, other than the remixes being, usually, great, is that Razormaid were the first to get permission and pay royalties to the artists (I could be wrong…); but it is a fact part of the reason their work is so good is that they got access to the master tapes; I know there are already lawsuits against certain companies for using certain artists’ data as training; and there are already, at least, claims by the awful, awful big record labels against AI generated songs.

I understand you probably want to share your enjoyment with the generated tracks with a like-minded audience and are, probably, not looking for profit; but I’m not entirely sure releasing them would be a great idea from a legal standpoint.

As the youths of today like to say… TLDR: release some great original music you created! :slight_smile:

7 Likes
5

Of course I will call it a rip-off, but in the end I don’t care.

I’m a human, I have dreams. I don’t think that would be disrespectful if you want to pay homage to them, just be clear that’s an homage based on something fun based.

I think “if the idea is good, the final product could be nice”. the first time I heard some AI music it was “NOFX’s Linoleum” inspired, something that still smells like my carefree youth days. And I will remeber it with a smile. But if NOFX still smells youth, AI smells like youth.

If you enjoy doing it, why not? :broccoli:

6

In the very near future, these companies will be AI-generating their own “new releases”,and only be suing others using their library. Like many large corporations, they love the idea of never having to pay humans for work ever again.

1 Like
7

I appreciate the lively and thoughtful debate. It seems like the consensus is:

  1. Creating musician-inspired music for your own listening: Probably OK
  2. Sharing it for free: Debatable
  3. Selling it: Not OK (which make sense)

I shouldn’t try to boil it down like that, but I like to reiterate things so that you know I’m listening. :slight_smile:

I have my own ideas on the subject, but I’m more interested in hearing yours. Mine aren’t anything new anyhow.

1 Like
8

That is sadly true, indeed.

One bit I forgot to mention above; but I have discussed with acquaintances before… there’s a certain thing that makes us human… a spark? A soul? Being connected with the pineal gland? Whatever people want to call it… it is part of what drives us to create. Cave paintings, clicking two sticks together with a certain tempo, deep sounds on the skin of some animal, complex machines with black and white keys to torment our fingers but release our ideas… computers, AIs… will never, ever achieve that.

3 Likes
9

A vast topic indeed! Which leads me to ask a question: isn’t the VCV Rack user essentially exploiting a kind of AI (consciously or unconsciously)? Particularly when using certain modules that rely on random or pre-programmed “musically” processes (I’m thinking specifically of random sequencing, or a harmony generator or melody generator). Because ultimately, by using them, the designer doesn’t truly have control over the final output of the “composition”…

10

By no means are stochastic processes AI.

Composing music using chance processes has a long history. Mozart experimented with composition using dice, and many more examples.

3 Likes
11

I really love music and I wish AI would just go away.

I do not want to consume this bullshit at all.

I am not alone, the majority of people I speak with about music and art are aghast at the notion that prompts are making music/art that is shared like actual art.

The AI bubble as it is being played out currently is bad for our culture and ourselves in too many ways to express it here.

2 Likes
12

There is a bit of truth in this for my modules. My modules are based on my work in commercial “expert systems” development in the late 80s. This was the state of the A.I. art in the late 80s. I first coded my plugin and modules as a DOS/Windows application in 1988.

1 Like
13

That is why I am always tempted now to describe it as aleatoric music. It’s actually a less accurate term, but the term generative is just thoroughly poisoned now, I feel.

The main problem here is probably that the vast majority of people have not even a coarse understanding of what those models do. It’s not only pretty complicated, but also actively obfuscated and additionally muddied by terminology. It’s a fine mess.

2 Likes
14

While I can see the point of the question; I don’t see stochastic methods being the same as AI at all: Stochastic processes are a composer’s decision and under the composer’s control.

Some modules can produce the same output over and over again when restarted, that could lead to “sameness” across different patches and composers; user control and choice, however, can lead to vastly different outputs from the same module; that said, the “sameness” part can also be made composer selectable or automated if the developer deems it so. For example, Sanguine Mutants Marmora can reseed itself using triggers and, also, using the context menu; the user can input a different, arbitrary, seed whenever desired and the upcoming version can seed itself when loading patches, etc. The vanilla Marbles port always uses 1.

AI, on the other hand, produces something from “Write an industrial song that sounds like Kant Kino”… and uses either parts of Kant Kino’s music or something that sounds similar, it also tends to produce finished songs… unlike using generator modules: they are only part of a whole that will need work and follows deliberate, concious intent.

1 Like
15

I think that’s the key term… it will eventually burst; it seems it already is beginning to…

1 Like
16

I know man, it’s supposed to be doing the dishes while I make music, now I have to do the dishes while its making music. Wtf??

5 Likes
17

And on top of that, fighting it is like farting against thunder, its pointless when artists are the minority. Heres the sales pitch: “now anyone can make music even if you have no clue what you are doing”. Well yes, thats why your song sounds exactly like the previous one, and you’ve learned or created nothing out of the process.

2 Likes
18

You may be right, but I’m not sure I would agree. Like with every new technology there will be drawbacks that sometimes will be significant, but I believe AI is here to stay (and to further develop) in the areas where it is perceived to be ‘superior’, such as being cheaper, more efficient, more effective etc.

Consider the development of the steam engine vs. muscle force or the pocket calculator vs. the occupational (human) computer. Also, I’d like to point out that the use of music goes far beyond just being an art form. I am convinced that for these mundane uses, such as advertising jingles or elevator music, AI will become dominant (if it isn’t already).

Not saying I like this development, but my crystal ball tells me this is where it is heading. :shaking_face:

1 Like
19

No. I know that most people haven’t got a clue about computers or software, and couldn’t tell an AI from a program if it hit them in the face. We’re already at the hype-stage where if something uses an application that does something useful it’s called “AI”. AI at the current moment means something specific: Large Language Models (LLM’s). They are highly sophisticated token-prediction-programs and work fundamentally different than other traditional programs.

No. When we talk about “generative” techniques we’re talking about introducing some level and form of randomnes into the musical process. Randomnes and AI are two completely different things. The outcome of using AI are very much supposed to be not-random.

Yeah, I’ve been devoting a fair amount of thought to technological development lately, and find myself becoming more of a luddite I’m afraid. I more and more tend to think that replacing human activity, even mundane things are mostly a mistake. Also, I think it’s a fact that we’re becoming dumber as our use of technology increases. As we farm out activities that before would occupy brain cycles, we gravitate our time spent towards entertainment or damaging forms of social media. I don’t think it’s a good development at all.

I tend to agree, unfortunately. I don’t think it contributes to human flourishing, on the contrary.

4 Likes
20

I also fear that those that speculate on a so-called “AI-bubble” are doing a kind of eschatology, especially those that are staunchly anti-AI in principle. If there is no “popping” of the AI bubble, and it does steadily hum along with the kind of mishaps and innovations that come along with any new technology, you may be putting yourself at a real disadvantage by refusing to use AI at all. If the day it bursts never comes, how soon will these hardline anti AI users have to concede some ground just by the mere fact that AI might someday be largely present in our every day life, just like computers and smartphones are today, for better or worse?

All the while, those that aren’t so dogmatic will at least be able to do the work of navigating the current jungle of products and tools to find what might be a workable status quo in the future. How this relates to music - lot’s of the arguments against AI projects that create sound and music presume that they will only be used to make entire finished and mixed songs generated from a dataset of completed music. It is implausible that it will remain the case for too long. There’s a certain pessimism that AI will only be used to create abject trash: jingles and muzak just like you said. This pessimism throws out the baby with the bathwater. Of course people who engage seriously with music won’t be impressed by elevator music, AI won’t be used to create just that! It is probably naive to think that eventually your favorite artists won’t be using AI at some point in the process in their craft. It is our job as artists and listeners to decide what is and isn’t acceptable within the realm of our craft.

With all of this said. I wouldn’t even call myself pro-AI in any meaningful sense. I’ve been trying to teach myself math via self-study. Among other resources, I’ve found that AI seems like a sensible solution to aid in this personal endeavor. Even if I did hire a tutor, there are certain sanity-checks you can conduct with an AI - who never runs out of time to answer your questions - that you just cannot with a real person. That said, I’ve been immensely frustrated with just how wrong AI often is. Being mislead about a technical field like math is incredibly frustrating, and I am on the fence about how feasible it is. Considering that, AI assisted sound creation and music has considerably lower stakes, yet I frequently see artists railing against AI much more heavily in this discipline than in technical ones. Perhaps the moral circumstances are the leading issue in this regard, but I fear it may be a losing battle. And if we must lose the battle, it may be possible to not lose the “war” so-to-speak if you at least engage minimally with the forefront of AI, and advocate at least for the ones that are actually…good and useful, assuming there could be such a thing!

2 Likes