Comment by lukevp
2 days ago
And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
AI music is the same as AI code. It’s derived from real code, but it’s not just regurgitated wholesale. You still as a person with taste have to guide it and provide inputs.
Electronic music made it so you didn’t have to learn to play an instrument. Auto tune made it so you didn’t have to learn how to sing on key. There are many innovations in music over time that make it easier and less gatekeepy to make music.
We are just moving from making music as a rote activity similar to code, to making music like a composer in much the way that you can create software without writing code. It’s moving things up a level. It’s how the steady march of innovation happens.
It won’t work to put the genie back in the bottle, now it’s to find what you love about it and makes it worth it for you and to focus on that part. Banning the new types of art is only going to last as long as it takes for people to get over their initial shock of it and for good products to start being produced with it.
>And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
Personally, I don't buy this "AI models are learning just like we do." It's an appeal to ignorance. Just because we don't fully understand how a human brain learns, one can't claim it's the same as a statistical model of ordered tokens.
But even if it were true, I'm alright with drawing a line between AI learning and human learning. The law and social conventions are for humans. I want the ability to learn from others and produce original works that show influences. If this right is allowed to all humans, there is a chance one learn from and outperform me. That would suck for me, but I can accept it because it came from a universal human right I also enjoy. But an AI model doesn't have human rights. For models, the law and social conventions should still favor humans. The impact on the creative community and future creative endeavors should be balanced against the people who create and use the models.
I don't know how to do that with LLMs in a way that doesn't prevent the development of these amazing models. Maybe the government should distribute a portion of the revenue generated by the models amongst all citizens, to reflect how each model's value came from the written works of those citizens.
> If this right is allowed to all humans, there is a chance one learn from and outperform me. That would suck for me,
This is a rather sad take. If someone learned from my art or music and did something new and more popular, I would be happy! I had influence, I mattered. That new more popular work takes nothing away from my previous work. In fact, when I do science I'm doing it explicitly for this reason, to build on.
For me, creating music is not about "being the best" or "making more money than some other artist." It's about telling the stories I want to tell. An AI would not tell my stories, ever. It might produce things that somewhat similar, but it won't tell a human story, just a shallow imitation.
On the flip side, AI can be immensely useful. For example, stemming means that DJs or visualizer applications can do more with music. Perhaps AI can be used to create interesting new effects, or interesting new instruments or sounds. It can give ideas and help with inspiration.
I honestly have a hard time seeing AI actually driving musicians out of business because it can't tell a story. And it can't do that because it hasn't lived a life. Yes, I can see it producing low quality ad-jingles or low quality filler tracks like you see in spotify, so some people will be impacted. But we're long past time for some form of universal basic income to deal with this. It's not just artists that need a basic income at this point.
You didn't finish the sentence:
>That would suck for me, but I can accept it because it came from a universal human right I also enjoy.
>The law and social conventions are for humans.
I don't know about that. America shows us that laws and social conventions are for corporations. Humans are just entities to extract profit from.
We don't talk about it much in these AI topics, but there's definitely the elephant in the room of the whole "low trust society" aspects that make a lot of actions automatically scrutinous from corporations, especially American.
But I've seen the discussion here on that's and we're pretty far away from being able to have a good discussion on that. Let along bridging the two topics together.
> Electronic music made it so you didn’t have to learn to play an instrument.
This is cliche. Most celebrated artists in the electronic music world can play several instruments, if not expertly, than at least with enough familiarity to understand the nuances of musical performance.
Electronic musicians are more akin to composers and probably have more in common with mathematicians and programmers in the way that they practice their craft, whereas musical performers probably have more in common with athletes in the way that they practice their craft.
You also need to understand how instruments make sound at an engineering level if you want to make timbre-perfect synthesizers which sound like said instrument, for instance
Electronic music is also very closely related to computer animation. Animated film technology is much more advanced, but a lot of techniques are similar.
Probably a good analogy too. Pixar's creative process is quite different from drawing it frame by frame and at least some aspects of it will use have used some sort of generative process, but it's incredibly involved and conscious in a way that typing "video of cute cartoon cat, Pixar style" into a prompt isn't.
Same applies to Bandcamp not having any issues with people making music in a DAW
3 replies →
> Electronic music made it so you didn’t have to learn to play an instrument. Auto tune made it so you didn’t have to learn how to sing on key.
Neither of those things are really true, though. They made it possible to make poor music without learning those things, I suppose, but not make good music.
> Banning the new types of art
Nobody is seriously talking about banning AI generated music. What you're seeing is a platform deciding that AI generated music isn't something that platform is into. There are a lot of different platforms out there.
What is "good" music?
Perhaps music that at least the author would listen to? To this day I haven't heard an AI song that made me wish I press the rewind/play to listen it again. Granted, most human-generated songs are crap, too, but at least they are not crap to their authors.
1 reply →
The eternal question.
I think in this context, the term "intentional music" or "earnest music" applies better. People who just wants "music that sounds good" already has mainstream stuff. Many who want a more niche sound deliberately look to support humans in that endeavor. Not yet another billionaire label who puts out "safe" but "boring" stuff. Except it's worse now.
Humans are humans, computer programs aren't. A computer program learning doesn't matter, and it's not comparable to human learning. I have no empathy, sympathy or any sort of allegiance to computer programs.
I would imagine the vast majority of other humans agree with me. I'm not just gonna betray humankind because some 1s and 0s "learned" how to write music. Who cares, it's silicon.
> And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
The volumes of production are really scales of magnitude of difference between a human producing music, and a computer.
With a script and generator 1 individual could oversaturate the whole marketplace overnight rendering it impossible for other individuals to be found let alone extract any value.
Also, I don't know if you've ever done music production for fun but you don't really just setup only a prompt. It takes a significant amount of time to actually produce something. Time setting up a DAW system and export an empty track, and submitting it. An empty track.
Let alone actually doing all the microoptimizations by ear and trial to produce any catchy tune. Meanwhile a statistical approach doesn't even have to understand what's it's doing, could as well be white noise for all it matters.
> AI music is the same as AI code. It’s derived from real code, but it’s not just regurgitated wholesale. You still as a person with taste have to guide it and provide inputs.
I guess the difference is proprietary code is mostly not used for training. It's going to be trained on code in the public. It's the inverse for music, where it's being trained on commercial work, not work that has been licensed freely.
LLMs are absolutely trained on commercial work. You just need to look at the lawsuits coming out against the AI companies.
> Or are you also extracting other artist’s work and using it as inspiration for what you do?
Yes, when I make music, I am taking inspiration from all of the other artists I've listened to and using that in my music. If someone listens to my music, they are getting some value from my contribution, but also indirectly from the musicians that inspired me.
The difference between that and AI is that I am a human being who deserves to live a life of dignity and artistic expression in a world that supports that while AI-generated music is the product of a mindless automaton that enriches billionaires who are actively building a world that makes it harder to live a life of stability, comfort, and dignity.
These are not the same thing any more than fucking a fleshlight is the same as being in a romantic relationship. The physical act may appear roughly the same, but the human experience, meaning behind it, and societal externalities are certainly not.
100%. I think there are some clear distinctions between AI training and human learning in practice that compound this. Humans learning requires individual investment and doesn't scale that efficiently. If someone invests the time to consume all of my published work and learn from it, I feel good about that. That feels like impact, especially if we interact and even more if I help them. They can perhaps reproduce anything I could've done, and that's cool.
If someone trains a machine on my work and it means you can get the benefit of my labor without knowing me, interacting with my work or understanding it, or really any effort beyond some GPUs, that feels bad. And, it's much more of a risk to me, if that means anything.
> If someone invests the time to consume all of my published work and learn from it, I feel good about that.
Agreed. My goal, my moral compass, is to live in a world populated by thriving happy people. I love teaching people new things and am happy to work hard to that end and sacrifice some amount of financial compensation. (For example, both of my books can be read online for free.)
I couldn't possibly care less about some giant matrix of floats sitting in a GPU somewhere getting tuned to better emulate some desired behavior. I simply have no moral imperative to enrich machines or their billionaire owners.
> I am a human being who deserves to live a life of dignity
Sure, but so does the homeless guy living on the streets right now because computers and the internet automated his job - and yet here you are using the very tools ("mindless automatons") that put him out of work.
That's a good observation, but it doesn't cancel out the GP's point, or its author's dignity. On the contrary, actually, it provides more depth and force to their argument.
A given technology may benefit some while harming others. And it may have harms and benefits that operate on different time scales.
The invention of the shipping container put nearly every stevedore out of a job. But it made it radically cheaper to ship things and that improved the quality of life of nearly everyone on Earth.
I suspect that for most stevedores, it was a job where the wages provided dignity and meaning in their life, but where the work itself wasn't that central to their identity. I hope that most were able to find other work that was equally dignified.
That's certainly less true for musicians, poets, and painters where what they do is central to the value of the work and not just how much they can get paid.
There's no blanket technology-independent answer here. You have to look at a technology and all of its consequences and try to figure out what's worth doing and what isn't.
I think shipping containers are a pretty clear win. I think machine learning for classification is likely a win.
It's not at all clear to me that using generative AI to produce media is a win. I suspect it is a very large loss for society as a whole. Automating bullshit drudgery is fine. Most people don't want to do that shit anyway. But automating away the very acts that people find most profoundly human seems the height of stupidity to me.
Do you really want to live in a world where more people have to be Uber drivers and fewer people get to make art? Do you want to live in that world when it appears that the main people who benefit are already billionaires?
4 replies →
Spot on Sir
[flagged]
> AI music is the same as AI code. It’s derived from real code, but it’s not just regurgitated wholesale. You still as a person with taste have to guide it and provide inputs.
Not necessarily apples-to-apples here. Full songs generated from AI prompts don't crash like a computer program would. You could simply upload the garbage to Spotify and reap the rewards until it got removed (if it even does).
Some of the worst (best?) AI "artists" on Spotify have millions of views. It's tragic what it says about us. That most of us not only can't tell, but actually prefer this kind of uni-tone, blase, on-the-nose, emotionally manipulative crap.
> That most of us not only can't tell, but actually prefer this kind of uni-tone, blase, on-the-nose, emotionally manipulative crap.
This is already what pop music, EDM and some other genres have been about for decades. Most of it is slop made with overused similar chord progressions and beats. The very fact that we can easily separate music into genres is a proof most of the music we produce nowadays is super generic and follows very basic repetitive patterns.
There is AI slop but there is human slop too and it tends to be very successful.
There's music and there's music. When I want to listen to Music then I pick an artist and album manually. But 99% of the time, I just need something to play in the background when I'm working or cooking or cleaning - then it just has to sound pleasant, the value of that for me is exactly zero. Some of the best mixes I find for that are ai generated because they have a uniform pleasant sound for a long time, without anyone trying to impart anything on them.
5 replies →
I think the analogy here is with Grok generating images of (real) people wearing bikini. It could always be done in Photoshop before (and with hand-made photo montages before that), but it's now accessible at scale to people with zero skill. That's when a quantitative change becomes qualitative.
Actually, to me this is the perfect argument to make AI music not have copyright.
Normally the copyright is owned by the creator. Algorithms can't own copyrights, so there is no copyright. There is already legal history on this.
> And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
For me, one key difference is that I can cite my stylistic influences and things I tried, while (to my knowledge) commercial musical generation models specifically avoid doing that, and most don't provide chord/lead sheets either -- I would find it genuinely sad to talk to a musician about their arrangement/composition choices, only to find they couldn't
> I would find it genuinely sad to talk to a musician about their arrangement/composition choices, only to find they couldn't
So much of music composition is what "feels right" and is instinctual. Artists aren't consciously aware of probably most of their influences. They can cite some of the most obvious ones, but the creative process is melding a thousand different vibes and sounds and sequences you've heard before, internalized, and joined into something new, in a way only your particular brain could.
Let music historians work on trying to cite and trace influences. That's not something artists need to worry about.
> They can cite some of the most obvious ones
Thus already doing much better than the average Suno producer
E: More seriously, this strikes me as a motte-and-bailey where "Artists cannot list every single influence they have or provide an explicit motivation for every single creative choice" is treated the same as "artists cannot list influences or justify creative choices at all"
I am 100% sure you can't cite all of them
Depends - how long do you have, and do you accept answers in CSV, Arrow or Parquet?
> It won’t work to put the genie back in the bottle
It's not about putting the genie back in the bottle, it's about helping folks realize that the vague smell of farts in the air IS the genie--and this particular genie only grants costly monkey paw wishes that ultimately do more harm to the world than good.
> And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
This is an argument that the AI should be allowed to benefit, not the person prompting it.
> less gatekeepy to make music
Is "gatekeepy" how we're referring to skill now? "Man I'd like to make a top-quality cabinet for my kitchen, lame how those skilled carpenters are gatekeeping that shit smh"
Gatekeepy to not like something that's not to your taste
> And you create music without ever having heard music before? Or are you also extracting other artist’s work and using it as inspiration for what you do?
But the parent poster is, presumably, human! Humans have the right to take inspiration like that from other humans (or machines)! Why do we seem so keen on granting machines the right to take from us? Are we not supposed to be their masters?
Couldn't you just as well say it's a human taking inspiration from other humans through a machine?
Only if the human is actually making the music. If a machine is just generating the song at a human's request, then the human isn't making music, the machine is.
No. Because the inspiration does not pass through the human, only through his machine.
>We are just moving from making music as a rote activity similar to code
From this statement, I doubt you've written any music worth listening to, or any code that's not trivial.
Don't confuse music with muzak. What you get from an "AI" is muzak. It will never, ever have the same depth, warmth, or meaning as a human translating human emotions and experience into music and lyrics.
Where did I claim in my post to have written music worth listening to or nontrivial code? Seems like you’re just insulting me in particular instead of providing a counter-argument.
There have already been AI-created #1 hits.
Sure, there’s music that has all of the attributes you lay out as “requisite” for “good” music, but this is classic moving of the goalposts. It’s how people always justify that AI is not here yet, because there’s this facet of it that’s not human enough.
A lot of the music people listen to is devoid of the depth, warmth and meaning you mention, even without AI involvement. It’s written and produced by tens or hundreds of people and there’s no single visionary behind it. It’s a product.
Similarly, there can be AI assisted music that has just as much depth, warmth and meaning as a human, BECAUSE a human is involved in the decision-making of that music.
Do you believe that if someone uses a sample, or uses a prebuilt drum loop, that their music automatically is bad? What level of assistance is acceptable? Where do you draw the line?
> There have already been AI-created #1 hits.
It's an old story, but it was a fabricated one.
The only reason this sort of tracks is that a lot of people today don't listen to music, they just put it on as background noise to drown out the silence. It seems to pay off for some producers, but I don't think there's big money there, or a real threat of replacing artists.
By and large, the general public has shown that they notice the vapidness, blandness, and incongruity of GenAI music, and don't much care for it apart from seeing it as an interesting curiosity.
>Where did I claim in my post to have written music worth listening to or nontrivial code? Seems like you’re just insulting me in particular instead of providing a counter-argument.
You didn't, and I never claimed that you did - I wrote that I doubt you have. If you had written non-trivial code, or written any music worth listening to, then I doubt you would have the same conclusions.
>A lot of the music people listen to is devoid of the depth, warmth and meaning you mention, even without AI involvement.
I agree, and it will be forgotten, and that's fine. Not every song is a winner. I guarantee that #1 AI generated hit will not be thought about a year after it comes out. Yes we're still listening to hits from the 1960s that real people created because they express human experience that isn't easily fabricated by a machine.
> lukevp 13 hours ago | parent | next [–]
Where did I claim in my post to have written music worth listening to or nontrivial code? Seems like you’re just insulting me in particular instead of providing a counter-argument. There have already been AI-created #1 hits.
Sure, there’s music that has all of the attributes you lay out as “requisite” for “good” music, but this is classic moving of the goalposts. It’s how people always justify that AI is not here yet, because there’s this facet of it that’s not human enough.
A lot of the music people listen to is devoid of the depth, warmth and meaning you mention, even without AI involvement. It’s written and produced by tens or hundreds of people and there’s no single visionary behind it. It’s a product.
>Similarly, there can be AI assisted music that has just as much depth, warmth and meaning as a human, BECAUSE a human is involved in the decision-making of that music.
AI-boosting nonsense
>Do you believe that if someone uses a sample, or uses a prebuilt drum loop, that their music automatically is bad?
Generally, yes. I abhor Kanye and his ilk. YMMV.
> Electronic music...
Your instrument is the computer and designing sound. You still have to have talent and musical ear to make this music.
[dead]