Menu
What's new
By:
Filters

"Generative AI is the greatest risk to the human creative class that has ever existed"

marius_dm said:
Making music with samples suddenly feels like ancient technology. And for all the TLDR folks I see in this thread speculating about UDIO et al using samples, I don't think popular "AI" music generator tools (UDIO, SUNO) use any sort of "samples" in VI-Control's understanding of the term. It just trains on spectral data (FFT) of existing music and then spits out new stuff based on that data it was trained on ("influences"). But could as well be trained on a bunch of MIDI files, so then it could spit out derivative MIDI music based on that training (AIVA et al). Or take it further and also have a model that's trained to output real sounding music from a MIDI file you provide (here's where it gets interesting for this audience).
Click to expand...
So, Training on FFT data is essentially like a super deep sampling ( but different workflow)

If I analyze ( and record ) an FFT of a sound so thoroughly and in-depth, to have all the structure a sort of DNA, it's pretty much the same as sampling, albeit through a different method. Therefore, this approach still potentially breaches ANY copyright rules ( if done on material covered by copyright, as it actually is done)

Sampling means to capture, to store information about something. It doesn't matter if it's Biology, an audio signal, or FFT data. It's all about sampling.

For instance, if I input some data in UDIO and it outputs EXACTLY Andrea Bocelli's voice, that's a blatant sign his vocal structure has been replicated. And frankly, that's not exactly legal.

If I sample Andrea Bocelli's voice, I have to pay Andrea Bocelli. I hope that's clear.

So I don't see a single valid reason why these people should be allowed to behave differently. The same rules should apply to everyone.

And considering the amount of data these people have analyzed, there's a potential risk that they've broken thousands of rules.

Not just with celebrities, but with everyone, even a simple developer like me or others. Perhaps many videos circulating on YouTube have been used as training material. And I must repeat, this is not acceptable.

Then you talk about making music with samples, but do you really think using UDIO means you’re making music? You're not actually doing anything...
 
Last edited:
Well of course it uses "samples", or full recordings rather, but like I've said, not in VIC's understanding of the term. I wasn't arguing the point in the context of copyright at all btw. I simply saw some earlier posts from folks speculating UDIO must be using some advanced sampling technique to get that realistic output, lol.
 
SAG-AFTRA and leading record labels reached a tentative multiyear agreement on a successor contract to the SAG-AFTRA National Code of Fair Practice for Sound Recordings.

Covering the period starting Jan. 1, 2021, and ending Dec. 31, 2026, the agreement includes Warner Music Group, Sony Music Entertainment, Universal Music Group and Disney Music Group.

The SAG-AFTRA Executive Committee unanimously approved the tentative agreement on Wednesday, April 10. It will now be sent to members for ratification.

thetrichordist.com

@SAGAFTRA and Major Labels Reach a New Sound Recording Agreement With AI Protections

SAG-AFTRA and labels reach a new deal with AI protections.
thetrichordist.com thetrichordist.com
 
For example:

A. I choose the best snare drum samples from ten sample developers and mix them to a single snare drum sample that I sell commercially.

B. At the same time AI uses the same ten snare drum samples, processing them by filling them with white noise which is subsequently removed.

In these examples, the result in A and B are 100% identical. Is one of the methods legal, appropriate and conforming with license agreement while the other one is not?
 
Alchemedia said:
SAG-AFTRA and leading record labels reached a tentative multiyear agreement on a successor contract to the SAG-AFTRA National Code of Fair Practice for Sound Recordings.

Covering the period starting Jan. 1, 2021, and ending Dec. 31, 2026, the agreement includes Warner Music Group, Sony Music Entertainment, Universal Music Group and Disney Music Group.

The SAG-AFTRA Executive Committee unanimously approved the tentative agreement on Wednesday, April 10. It will now be sent to members for ratification.

thetrichordist.com

@SAGAFTRA and Major Labels Reach a New Sound Recording Agreement With AI Protections

SAG-AFTRA and labels reach a new deal with AI protections.
thetrichordist.com thetrichordist.com
Click to expand...

Nice,

but

This helps all the major artists and singers under contract with multinationals. Unfortunately for independents, there is no protection.

And only by boycotting these systems might we be able to slow down the disaster somewhat.
 
This lecture (as part of an overall course) gives a good overview of how these models works and the current state of the AI music industry, particularly from 13:40 onwards,



My main takeaways are this tech has been around for some time but only now we are getting the compute available to make it practical to a mass market, (its interesting to see the development over the years and different technologies to this point)

Current models are trained on large amounts of audio rather than samples or using midi,

It was also very interesting to me how Mozart had the foresight to help invent the first generative music, so the idea itself is not a new thing
 
Brendon V said:
This lecture (as part of an overall course) gives a good overview of how these models works and the current state of the AI music industry, particularly from 13:40 onwards,



My main takeaways are this tech has been around for some time but only now we are getting the compute available to make it practical to a mass market, (its interesting to see the development over the years and different technologies to this point)

Current models are trained on large amounts of audio rather than samples or using midi,

It was also very interesting to me how Mozart had the foresight to help invent the first generative music, so the idea itself is not a new thing
Click to expand...

Sweden was a leading nation in piracy 15 years ago or so and this was justified in state media over and over. The Pirate Bay owners were hailed as heros, and the bottom line propagated was that, just like anything else in the welfare system free computer games, movies, music, were a human right. Swedish National Radio had multi-episode series "explaining" that there was no original music - only derivative music, and artists, musicians, and composers were only carrying on a tradition, and they had no rights to royalties for their music because it wasn't theirs in the first place. They were just re-hashing what belonged to everyone and therefore piracy of music without compensation to the artist was justified and righteous. I only watched the first five minutes of this video pod, but to me, it oozes of the same justification for leaving the creators without compensation: already Mozart's music was just s product of random algorithms. Without having seen the whole half hour, I get a really bad taste in my mouth from the first five minutes.
 
FinGael said:
I am already getting stomach pain from the mental image of 1463 different Marvel sequels and prequels.
Click to expand...
I guess in a few years, people will be able to consume as many AI-generated Marvel movies as they want. As will everyone else be able to enjoy endless amounts of whatever they desire. Perhaps it will mean that "hit movies" won't be a thing anymore, and so you won't have to suffer endless marketing of movies you don't want, because everyone will be living in their own personal content heaven. Strange times ahead.
FinGael said:
Several prompters needed. "Hey man, at least it's a job". Sure, until the AI does the prompting too.
Click to expand...
I might be wrong, but I'm guessing that people using the AI tools now are largely unaware that they're bringing about the exact situation you are describing
 
Last edited:
Today’s NY Times Daily podcast is about an investigation into the types of data used in AI models, including the legal grey areas. Interesting listen.

EDIT: here is the gist… All of the big AI companies have been forging ahead, training on copyrighted material, arguing fair use because what they do is transformative, but that it’s a legal grey area the courts are going to have to decide on and that to keep improving the models, they will likely need to keep sucking up copyrighted data.
 
Last edited:
If this gets me closer to the Replicator from Star Trek, and living forever, I'm all for it!

But seriously I'm quite impressed with the recent developments. As a composer I know I'll be put in my place by AI soon, if not already; but as a pianist I can't wait to tell it to print me the charts so I can play whatever interesting things it will come up with.
 
Hans Adamson said:
For example:

A. I choose the best snare drum samples from ten sample developers and mix them to a single snare drum sample that I sell commercially.

B. At the same time AI uses the same ten snare drum samples, processing them by filling them with white noise which is subsequently removed.

In these examples, the result in A and B are 100% identical. Is one of the methods legal, appropriate and conforming with license agreement while the other one is not?
Click to expand...
Are you allowed to do that? I would have thought sample libraries have a policy which says you’re not allowed to re-sell the samples as samples.

i.e I can’t just mix 6 Infinite Brass horns together and sell it as a horn ensemble patch
 
The main question is becoming more and more about what is relevant and how to make anything relevant. Before AI the question was both how can you create anything relevant and then how can you actualise that relevancy in reality. Now the creation side is just endless stream of stuff, both human creations and AI creations. But in some ways the value of that created whatever starts to head to zero, because just existing doesn't mean anything. When everything exists, nothing has any value in itself.

Now the question is how can we filter the relevant stuff from pointless, and that is both marketing issue and cultural issue. Where do we find relevancy in daily lives? Do we lose the ability to filter the non relevant from relevant? I personally have found that in the past decade I've watched less and less movies and tv-series, because while the offering has become wide and vast, the relevancy for me personally has plummeted. I realise that I mostly go back to watching the 90s era films and series like Northern Exposure, The West Wing and such, that were creations of totally different culture for totally different audience.

I'm afraid AI generated endless possibility can make us as consumers numb, because suddenly we find that while we can create and watch anything, it probably makes us feel nothing.

Maybe the human condition is that we need limitations because otherwise we don't know how to be
 
SamC said:
This is the reason the film and tv industry will be reticent to touch AI music. Film/TV/VG Studios own the music created for them. If they can’t own it and continue to commercially exploit it, why use it.
Click to expand...
Said this elsewhere but consider this: if it costs them very little to generate it’s a net win for them, because they own the stream. If they make no claim on the music they can still own every other aspect. And if AI were good enough to generate the whole thing they wouldn’t have the production expenses and could just profit from the subscription.
 
I don't get the fear of AI, just go to Apple Music or Spotify & just click a random button. A click can playback the whole song and you can repeat that for millions of times. It's not going to be Brahm's Piano Concerto No. 2 every time. But you clicked a button and there's a song. Who cares if it's generated by AI or by humans. A song is a song. Not like AI is generating stuff like Wagner or Beethoven or even Beach Boys.

Even you who fear AI are not creating anything that compares with Brahms or Schubert or Velvet Underground. I want AI to be so good that it eliminates all mediocrity. There's no Wagner before Wagner, no Beethoven before Beethoven or No My Bloody Valentine before MBV. These fuckers are creating shitty stuff that's just a bad copy of stuff that came before. AI is never gonna create anything that'll be like it's AI and anything after we'll say it's AI. AI just works on a sample data and creates similarity. We never cared for similarity. It's not artistic. Tarkovsky would've removed a scene if it resembled the style of Bresson or Felini or Kurosawa or Bergman. Yet, these people just copies and reproduces a bad replica and call themselves artists. I wouldn't care if they die. I even wish for it. Bloody Artists, my ass. It was never about Money. Render greatness to God & let Cesar give you what he deems right. An Artist's obligation is to God and not to Cesar

It's the same fear which is just a procrastination that's no different than to people who believe in different timelines or conspiracy theories or just some dumb stuff. Just for the sake of avoiding the look in the Mirror. "Of how much coward I am to not even try cause that'll hint the potential and I can't even accept that my potential is not that much. I much prefer too never know that potential cause it's better to be ignorant than realize my dispensability" They'll never want this thought in their lives. They'll just care about dumb things and create dumb stuff. For God's Sake enough of these whores.

Prostitution is better than this shit. At least while being a prostitute you're sacrificing you're body for money which god will forgive. It's even graver to prostitute Art which is like prostituting god himself.

P.S. - This should offend no one. If the language gets a bit crude, forgive me.
 
Yashiro_Ryu said:
I don't get the fear of AI, just go to Apple Music or Spotify & just click a random button. A click can playback the whole song and you can repeat that for millions of times.
Click to expand...
It's why AI music was not needed. Consumers already have too many choices of music created by humans.

Spotify is also good at identifying your tastes and will already feed you with a seemingly infinite number of pieces that you will probably like.

AI music only helps the big guys and tech companies. Consumers don't even really benefit from it other than some short term fun with the toy aspect of it.
 
Sfpx said:
It's why AI music was not needed. Consumers already have too many choices of music created by humans.
Click to expand...
This, and it’s why I see music AI as mostly pointless. How many people go searching for random music to listen to? Maybe it’s more than I think, but it seems so bizarre.

I can see a market in AI library music where producers are looking for a certain kind of thing that passes a content ID check, more or less hits an emotional cue, and is inexpensive. It’s like using MidJourney to illustrate a personal blog post. There’s a real market here but I doubt that it’s all that large.

There’s likely a market for licensed soundalikes. The Beatles singing your lyrics for $5, commercial exploitation prohibited. Cool. Commercial exploitation for a hefty additional fee. I can see a market in that.

Beyond that though? Not really.
 
Yashiro_Ryu said:
I don't get the fear of AI, just go to Apple Music or Spotify & just click a random button. A click can playback the whole song and you can repeat that for millions of times. It's not going to be Brahm's Piano Concerto No. 2 every time. But you clicked a button and there's a song. Who cares if it's generated by AI or by humans. A song is a song. Not like AI is generating stuff like Wagner or Beethoven or even Beach Boys.

Even you who fear AI are not creating anything that compares with Brahms or Schubert or Velvet Underground. I want AI to be so good that it eliminates all mediocrity. There's no Wagner before Wagner, no Beethoven before Beethoven or No My Bloody Valentine before MBV. These fuckers are creating shitty stuff that's just a bad copy of stuff that came before. AI is never gonna create anything that'll be like it's AI and anything after we'll say it's AI. AI just works on a sample data and creates similarity. We never cared for similarity. It's not artistic. Tarkovsky would've removed a scene if it resembled the style of Bresson or Felini or Kurosawa or Bergman. Yet, these people just copies and reproduces a bad replica and call themselves artists. I wouldn't care if they die. I even wish for it. Bloody Artists, my ass. It was never about Money. Render greatness to God & let Cesar give you what he deems right. An Artist's obligation is to God and not to Cesar

It's the same fear which is just a procrastination that's no different than to people who believe in different timelines or conspiracy theories or just some dumb stuff. Just for the sake of avoiding the look in the Mirror. "Of how much coward I am to not even try cause that'll hint the potential and I can't even accept that my potential is not that much. I much prefer too never know that potential cause it's better to be ignorant than realize my dispensability" They'll never want this thought in their lives. They'll just care about dumb things and create dumb stuff. For God's Sake enough of these whores.

Prostitution is better than this shit. At least while being a prostitute you're sacrificing you're body for money which god will forgive. It's even graver to prostitute Art which is like prostituting god himself.

P.S. - This should offend no one. If the language gets a bit crude, forgive me.
Click to expand...
<Hot take> Everybody is a copycat, some just hide and obfuscate it better and copy from a wider variety of sources than others. Creativity just means you've forgot where you're copying stuff from. </Hot take>
 
You must log in or register to reply here.
Top Bottom