Real Future Of AI in Music: Production Copilot Coming To Your DAW?

24 Likes

I don’t worry so much about “the art.” Lazy, safe, uncreative music already exists now, no AI required. People who get new tools find creative ways to push those tools.

My worry is about who benefits and who gets kicked to the curb.

7 Likes

All this i hear about AI in media is making me wonder why some people do art in a first place. I see it “useful” in marketing department, commercials etc. That was the only place left for people paying for “art”. But Art itself is safe , i love art, and have read many books about artists, it is to much pain and love and sacrifice involved in making of Art , but not much money… Above all, you don’t do art for money and to survive, you do art to sacrifice your self . Everything else is just missed opportunity.

9 Likes

that was right on the point :clap:
I will just add, all this media attention and space AI is getting lately , is not cheap, somebody pays good amount of money for it, and the question is why? and the answer is the real “news”. There is noting new about AI .

2 Likes

Honestly, having very few friends in real life who are into the same music I am, this would be an incredibly valuable tool for me.

3 Likes

I struggle to finish songs. I have no formal theory training. I make a lot of nice 1min “tracks”.
If an AI could point me in the right direction for arrangement/ finishing a song it would be awesome.
Like having a knowledgeable friend to work with.

3 Likes

reminder that a LLM will only regurgitate and combine what it’s been trained on.

so. a real risk of music treading water for a few decades if everyone defaults to these tools.
the innovation in the process will mean a lack of innovation in any output. me suspects.

was considering this last july:

my favourite recorded music has mistakes, noise, hiss and dirt.
i’m interested in how an artificial output will emulate human error and/or if it removes will it be great or boring as hell?

2 Likes

I was about to say that I loathe assisted creation, then remembered I own a Microcosm that kind of do something alike in a way.
And Mutable Instrument Marbles is fun.
Are these mostly smart randomizers, or an idea of what could be IA-assisted music creation?

Damn, I should create guitar pedals with some AI inside, just to try… Could be fun.

3 Likes

If you are trying to go for a certain sound (and there are many existing examples), using ML to get there faster seems like a good thing.

I wouldn’t look to ML to give innovative new ideas - but then, how often does music (or art generally) give us new ideas in the first place (that we are ready to experience)?

1 Like

I don’t mind using ML for some functions, like visual upscaling.

The more generative it goes, the less useful it is for me.

I’ve played around with Hookpad, it’s fine.

GPT and similar models have been pretty crapshoots beyond mildly amusing tricks.

If i could train my own models as easily as i do markov chain work for cut-up technique, perhaps it’d be more interesting.

But the general purpose just spouts bullshit or lies stone-faced and it’s more a timesink than it is helpful or a creative tool.

3 Likes

The problem is not “the art” so much as The Content (in an attention economy curated by algorithm) since several lifetimes worth of energy can be expended and automated, i’m not going to see human content as much on any social media platform. And i’d always prefer to see
even the most derivative human content over parlor tricks that’ll be using the same dark patterns as this soon enough-

5 Likes

This is exactly what I’m interested in. If I’m working in a genre outside of my expertise, having a tool to suggest edits, arrangement decisions, etc could be very valuable. As a productivity enhancer for people that want to finish things with a specific vision, it could be a great resource.

I share this worry as well. I like to think that it will mostly affect corporate music, adverts, etc, which I’m not worried about much. However, looking at the last decade or so, and seeing algorithm-driven trends in media grow ( 4 hour lofi hip-hop playlists to study and relax to, old jazz on YouTube suddenly getting millions of views, Spotify playlist-chasing artists, etc), I can’t help but worry that an AI revolution in music could further homogenize things. Although, now that I think about it, that seems to be inevitable anyway given the way that the Internet currently works and how companies use algorithms. AI might just speed it up.

1 Like

Yep! I don’t have a problem with actual targeted tools being incorporated, but it’s not how “AI” is being marketed or evangelized.

Anyone remember the “Half Life 2 mod: Seeks art, programmer, sound, level designer, community manager” ads people would place where they wanted to “coordinate” everyone with talent but not really create anything themselves?

I feel like this stuff promises people with no faith in their own creativity that they’ll be unblocked, but the ones seeking in good faith will not be unblocked, but the ones who just want to be “content creators” will be spamflooding their “works” everywhere.

Basically the people who don’t need ML integral to their workflow will be able to use it in small amounts, but the people most excited for this and with no concerns will either be disappointed or just shit out lifetimes worth of product.

Exactly, the video I posted is not explicitly about ML/AI, but it is absolutely pertinent to the discussion.

4 Likes

One of the issues with copilot is MS just appropriated code and unilaterally decided they had the right to make derivative works from it, ignoring the licensing. They seem to be safe, legally (yet to be seen, but they’re too rich to lose), but there is some precedent that 100% ML-produced content isn’t copywritable already in case law. That situation is likely to continue evolve, I expect ultimately with legislation.

I don’t think they would have found themselves in such a safe position with regards to melody. There are too many infringement rulings over melody where the chief complaint is roughly “I used some notes, and your song has notes in it too.” The music lobby is too massive. The code lobby is doesn’t really exist. SCO tried to sue everyone over nothing and lost. Oracle tried to sue Google over nothing and lost.

Sampling? That topic makes my blood boil, but I’m sure you’re all aware of the general situation.

On the creative side, I’m not interested. A music theory oracle and analysis tool would be nice, but only as separate tools. Sound design tools could be too good not to use, but by the time these things start spitting out quality audio, the next phase will be too close for it to matter. Synth programming tools, that’s an interesting middle ground, and well below the complexity of LLMs.


Ultimately, it’s gonna be machine radio. End-to-end automation of music production. You tell the thing what you want to hear, tell the thing to tweak the music as it plays if you like, and whatever. Prompt engineering here will be as common as playlist building. Good thing OpenAI is keeping things closed, otherwise the people might be able to run this stuff on their own, and then how could you charge for streaming?

But there are countless artists who get more attention than I do. I don’t think this really needs to change music production. If you’re in it for the money, idk what to say. Go to hell? Okay, if you love it and it happens to be your job: good for you. If we make it to post-scarcity, it can still be your “job.” If we don’t, well, losing the job of music to the machines will be pretty far down the list of complaints. I’m sure live music will stick around forever. At least until the humans go extinct.

(Ok, it might not be as all-or-nothing as I describe here, but if machine learning intensifies existing power structures, it’s gonna be a bad time. OpenAI is doing their best to protect the existing power structures. I’m sure they have convinced themselves its for good reasons, like someone might make a computer say a fuck word. Others too, but they went back on their mandate and are especially deserving of criticism.)

Regardless, I imagine there will still be an industry for artists, selected by the industry and sold to the public much in the way the big acts are now. We want that cultural point of reference, and we’re happy to pay for it. The music might be made by machines, but we still need humans in that part of the loop. We need them to be people. Or at least we need to believe it. I’m not sure things are going to really change much at that level.

I don’t know where musical genius comes in. Magnus Carlsen still blows minds even if a chess engine can make him look silly. I think music and chess are different here, where the musical can have value independent of the authorship. Engine chess can be interesting, but not something you’d get invested in.

There are some songs that I would lose all interest in if it turned out a machine had made them. Take Maggot Brain — that’s only gripping because something that was felt has been communicated in a way words could never manage. You can’t “prompt” an ML model to do that. I mean, unless it also feels (that’s coming too.) Hopefully the market for this kind of thing is strong enough that “real” music has a permanent and prominent place.

But I’m not all that negative about prompted music. Refining by prompting isn’t that different than refining by other means. The art isn’t in the labour, it’s in the artist’s senses. The only thing that really bothers me about this is it opens the door for the less passionate to “produce music.” Before I even consider whether that kind of gatekeeping is toxic or not: its moot. I cannot hear the passion in most of the music presented to me as it is. Maybe it’s a failure of understanding, but the point is manning the gate is not worth the effort, so you won’t find me on the corner with a “BOYCOTT BOOPSTEP” sign.

7 Likes

that was a really good watch, thanks for sharing that.:+1:t3:

I don’t like this sort of sneering attitude that everyone who wants to use AI, Crypto or whatever bollocks is somehow “forward thinking” while those of us who prefer to do things a different way are some sort of cave dwelling Luddite.

I mean, if I insinuated that all you guys pushing the “benefits” of AI were only doing so in the vain hope that our future Robot Overlords will look kindly on you and make you their house bitch instead of carting you off to the battery farms with the rest of us, you’d all be (quite rightly) offended. I’m not going to insinuate that though, because I’m not a prick.

Well I am a prick, I’m a prick that’s perfectly happy doing things without asking some Robot Gosplan knock-off whether or not my fucking kick is in tune. “Productivity tools” are for work, not real life.

11 Likes

Hey! That’s what my in-laws call me

1 Like

as a developer who is getting so much mileage out of GitHub Copilot, I’m super into any sort of DAW Copilot

2 Likes

There’s about to be an AI assistance tool for everyone with a desire to do what they want without much input.
sales co-pilot (super GAS)
political co-pilots
boredom co-pilots
friendship assistants etc.
It’s going to be wild on a level we’ve never imagined.
Again, it’s already past the point of no return, we just have to ride the pink cloud to the ground.

Or you could just turn your TV/phone/tablet/laptop/VR headset/smart assistant/smart doorbell/car/friendship group/microchip off and go outside and look at some trees for a bit.

6 Likes