• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

AI and Future of Music Production

It used to be ChatGPT never credited sources as it mashed together public info for a response. Now it does, due to a zillion copyright infringment threats.

The exact same should happen to music - we know existing generative AI models are based on "innovation through reassembly", it is just a matter of time until they are forced to publish which music/songs/composers were leveraged in generating AI music... and the originals get credited and compensated.
...all of them?
 
...all of them?
More specific? Do you mean all sources? I would say yes, and technically it is not that challenging. ChatGPT already does it by itself if you click on the little links... but I always ask for it as part of the prompt on top.

I have never tried to prompt to compose a song, but if I did I'd tell it to "write a mid tempo song in the style of Jimmy Jam and Terry Lewis that celebrates my mom's 80th" so right there..
 
More specific?
I mean that the source is...everything.

And since you are creating a piece of art, it's different than looking up facts where you can cite specific sources for the response. Laws are written such that reinterpretations of music are not copyright infringement...so compensation would be a stretch. For better or worse.
 
Last edited:
I saw a cartoon in the humor thread about a chicken crossing the road. I started writing funny lyrics, but my mind doesn't work like that so it came out a little more serious. These lyrics came to me really fast and after trying a few styles for the music and vocals I got this version on the 13th generation. It normally takes a lot more than that. I really like this melody and style.

 
Year Million takes some very interesting and perhaps disturbing looks at AI in the future. A really great National Geographic six-part documentary and science fiction series. Well worth watching!

Had a quick look - going to watch, thanks for the suggestion.

Though I think the timeline is way out. The world is still recognisable. People still living like we live - travelling like we travel etc but with super tech. It should probably be called year 3000 or perhaps 5000 (this is based on the opening of the first episode - perhaps later episodes are further out?)

998 thousand yeas from now - if humanity still exists, and technology has continued to develop close to current rates - it won't be recognisable to us as humanity. Hell - if AI gets it's way it won't be recognisable in 5K years.


Edit: having now watched the first episode - seems the series agrees with me. Year Million not a date - just the future etc.
 
Last edited:
I saw a cartoon in the humor thread about a chicken crossing the road. I started writing funny lyrics, but my mind doesn't work like that so it came out a little more serious. These lyrics came to me really fast and after trying a few styles for the music and vocals I got this version on the 13th generation. It normally takes a lot more than that. I really like this melody and style.


"The line is so yellow and the warning is seen
between the waking world and my poultry dream"

Genius. Pure poetry. :D
 
I mean that the source is...everything.

And since you are creating a piece of art, it's different than looking up facts where you can cite specific sources for the response. Laws are written such that reinterpretations of music are not copyright infringement...so compensation would be a stretch. For better or worse.
Interpretation of the law has changed over the years, with inspiration/interpretation that used to be acceptable now (and sometimes retrospectively) resulting in large liabilities. Under current interpretation James Brown might have been remembered more for copyright violation than innovation. Of course it still helps to be 'inspired' by someone who can't afford to sue...
 
I saw a cartoon in the humor thread about a chicken crossing the road. I started writing funny lyrics, but my mind doesn't work like that so it came out a little more serious. These lyrics came to me really fast and after trying a few styles for the music and vocals I got this version on the 13th generation. It normally takes a lot more than that. I really like this melody and style.

I thought the song idea seemed more of a folk-themed project.

 
It used to be ChatGPT never credited sources as it mashed together public info for a response. Now it does, due to a zillion copyright infringment threats.

The exact same should happen to music - we know existing generative AI models are based on "innovation through reassembly", it is just a matter of time until they are forced to publish which music/songs/composers were leveraged in generating AI music... and the originals get credited and compensated.

In practice, as long as the source samples are licensed (which involves compensating various parties, the last of which is the artist :-)), there is no need to credit the source. You see this in modern music: each song does not contain credits too all the samples it licensed from other artists. Modern commercial music generation platforms already license high-quality samples (e.g., Meta, Udio, Suno, ...), which is why they exist in the first place (the music industry is super litiguous).

For personal and non-commercial use, however, none of this matters.
 
In practice, as long as the source samples are licensed (which involves compensating various parties, the last of which is the artist :-)), there is no need to credit the source. You see this in modern music: each song does not contain credits too all the samples it licensed from other artists. Modern commercial music generation platforms already license high-quality samples (e.g., Meta, Udio, Suno, ...), which is why they exist in the first place (the music industry is super litiguous).

For personal and non-commercial use, however, none of this matters.

When one sees all the lawsuits between artists, I am sure things are not as straightforward. The limits of "fair use" seem to be established under litigation too often for us to assume things are as simple as that.
 
When one sees all the lawsuits between artists, I am sure things are not as straightforward. The limits of "fair use" seem to be established under litigation too often for us to assume things are as simple as that.
How do you or can you sue a computer for theft of "intellectual property" ? :p :facepalm:
 
How do you or can you sue a computer for theft of "intellectual property" ? :p :facepalm:
You don't. You sue the company that used the copyright material to train the AI. There are a number of cases winding their way through the courts, or getting settled between the parties to avoid setting a precedent.
 
You sue the company.
Yea, the courts all over the world are going to get piled up with this stuff for decades.
Me? I'll be looking down from Heaven laughing at it all soon enough. :eek:

GgOiL86WkAEYkZ-.jpg
 
For the open source and free AceStep 1.5 AI Music models:

The developers built the architecture to integrate diffusion-based generation with lightweight linear transformers and Deep Compression AutoEncoders. For copyright and training data sources, the developers state the models were trained on legally compliant, licensed, and royalty-free data
 
Watched this interesting short video from MIT about the 12 possible endings for a world with AI, based on a book by Professor Max Tegmark.

It's fascinating stuff . Seems like we should enjoy our noodling with AI generated music whilst we are still in charge!

 
Back
Top Bottom