Regarding Recent Technology Advancements

[jnWake]

55 minutes ago, djpretzel said:

If it's such a "completely different" technology than what the OP is referring to, take a look at: https://www.suno.wiki/faq/metatags/voice-tags/ - among other pages. These services are developing their own prompt tagging & nomenclature and some of them ALREADY let you provide an existing melody or sound clip that they will ingest & incorporate. The tech is ALREADY substantially beyond "a single basic prompt," and thus the OP is ALREADY insufficiently precise in what it is trying to communicate... yes, that's three capitalized alreadies, and one of them is bold AND underlined...

I mean, OP doesn't mention "basic prompts" or anything of the sort, it just mentions prompts. The "simple" part was by me 😋

There's really 2 points in the OP, one is about the ethics of how the models were trained regarding artist consent and copyright. I'm not sure if OCR's stance would be the same if the models were "ethically" trained or if, for example, a user trained their own model only on songs they made or something like that.

There's a second point about the "interpretation" aspect of AI generated music, the "human touch". I think we simply disagree here, I don't think current tools provide much "human touch". Even if Suno allows for more than "a simple prompt", the "human part" is still a low percentage of the finished creation. Of course this could change with time and how the technology evolves but it's my stance on what I've seen of current technology.

In the end, it's up to the community to decide how it approaches these things. I think it's fair if OCR decides it wants to focus on music that is created on a larger percentage by the "human touch" and, of course, policies can change over time as technology evolves.

[djpretzel]

2 hours ago, jnWake said:

I mean, OP doesn't mention "basic prompts" or anything of the sort, it just mentions prompts. The "simple" part was by me 😋

You're right - OP fails to clarify at all, and mentions "a prompt" - singular, not "prompts" - in its opening, bold line.

This is exactly what I'm saying is unclear; not specific enough.

2 hours ago, jnWake said:

There's really 2 points in the OP, one is about the ethics of how the models were trained regarding artist consent and copyright. I'm not sure if OCR's stance would be the same if the models were "ethically" trained or if, for example, a user trained their own model only on songs they made or something like that.

Right, that's ALSO unclear, because the waters have now been muddied with the ethicality/shade angle... which I don't think makes sense to even discuss until OCR has its own policy on whether OC ReMixes can be part of training sets, themselves... thoughts on that?

2 hours ago, jnWake said:

There's a second point about the "interpretation" aspect of AI generated music, the "human touch". I think we simply disagree here, I don't think current tools provide much "human touch". Even if Suno allows for more than "a simple prompt", the "human part" is still a low percentage of the finished creation. Of course this could change with time and how the technology evolves but it's my stance on what I've seen of current technology.

I don't get the feeling that either of us is entirely up to speed on the current tools, and hell, they're adding features on a daily basis. It's one thing to ban something and then say "Until such time as..." and explain how the tools would need to evolve... but that wasn't done, right?

Even with the current, limited tools, I can see someone spending a few hours with multiple prompts & revisions, tweaks at specific time markers, and some form of basic melodic input as part of the prompts, and then generating something that met most of OCR's OTHER criteria and ALSO involved more human input - more actual decision-making & human touch - than at least SOME mixes already accepted by the panel. The OP also says "wholly or in-part," so hypothetically one could generate a simple chord progression/structure using genAI and then add gorgeous, original, human vocals & instrumental performances on top, Made By Real Homo Sapiens!™, and it still wouldn't be kosher...

Is that how MOST folks are using the current tech? No... they're doing simple prompts, just as you say. Is it a good idea to ban a tool based on how most people might be "abusing" it? Then go ahead and remove the torrents, right? Otherwise, get to the real crux of what you're actually trying to avoid... the underlying principle.

2 hours ago, jnWake said:

In the end, it's up to the community to decide how it approaches these things. I think it's fair if OCR decides it wants to focus on music that is created on a larger percentage by the "human touch" and, of course, policies can change over time as technology evolves.

Well I wasn't there for the conversation, and for all I know Shariq decided this himself... who can say? Did the community decide already, or is it still deciding? It's a great, relevant, & essential convo to have; it's a little disappointing when a community's very first official statement about a whole new field/technology is to ban it with language that is unspecific and already - relevant to the state-of-the-art - outmoded.

What would that "larger percentage" of human touch be?

At any rate, regarding percentages, I'm 100% on board with the notion that the amount of human input & decision-making for submissions to the site remain a focal point, in perpetuity...

[Nase]

So, i'd honestly consider reviving PPR.

Well, we can Post nagging PPR questions in GD for now i Guess. ...

Can we? there's everything to Say today...more than ever. And reddit's short term mobile consumption structure Is so Bad For deep discussion.

That's what a forum Like This was good For at its best.

Maybe call It "forum americanum" or "Deep forum" or w/e...

I mean, maybe interest for This grows again and then DS reconsiders. For now, it's too empty Here. 

[Gario]

On 7/26/2024 at 11:04 PM, Nase said:

So, i'd honestly consider reviving PPR.

Well, we can Post nagging PPR questions in GD for now i Guess. ...

As long as you can keep the politics and philosophy relevant to video games and/or music, go for it. Coming from an old regular of the PPR forums, I don't believe OCR is a good place for off topic political discussion anymore, both because of the number of regular forum users (you need a lot of regular interaction to balance the topics discussed) and the effects some of the posts periodically have on those that do still use the boards, so I disagree with bringing it back to the site.

PPR was fun while it lasted, it can remain comfortably in our memory. In the meantime we should stick with discussing AI and music standards, as was the intent of this thread.

Edited August 1 by Gario

[Draconiator]

Another example.  Windchimes.  Those have no human involvement at all, so according to what I've been hearing on this thread, it'll be frowned upon to simply record windchimes and release it as some kind of new age album, simply because you merely recorded them?   Same thing here, if a bit different example.   I know that this example doesn't deal with human recorded works, but one could argue that if a passage in that album SOUNDS LIKE another person's work, your CD/album could get shot down.  "'Chimes of Destiny' sounds like 'a certain song'.  I'm suing you."   I think this legal challenge has no merit, on the basis of fair use.  They train on it sure, but they don't KEEP the source training.

Edit:  Another thing I forgot about.   This song I've attached was just made by Udio.  BUT NOT BY ITSELF.   I had to listen to 32 second chunks and actually CHOOSE which I wanted to include with the song.   The lyrics were generated by Claude, and even then I had to copy it, and paste the lyrics into the interface.  It took like an hour, so there is still very much the human element involved.   I'm sorry if anyone here hates country music....hehe

 

[DarkeSword]

"Recording wind chimes" is not in the same space as using machine-learning algorithms trained on consentless work.

The first post of this thread also says don't share work generated generated wholly or in-part by UDIO. Come on man.

[Draconiator]

Well if I can't share, lets do this another way.  Whatever happened to appreciating music in all its forms?  "OverClocked ReMix is a community dedicated to the appreciation and promotion of video game music as an art form."  But does that not extend to EVERY form of music, either human or machine? Why are we segregating music now?

 

I found this on the subject, and it goes over everything A.I.   I think AI music is still in Stage 2 according to this.  https://www2.datainnovation.org/2023-ai-panic-cycle.pdf

[DarkeSword]

28 minutes ago, Draconiator said:

Well if I can't share, lets do this another way.  Whatever happened to appreciating music in all its forms?  "OverClocked ReMix is a community dedicated to the appreciation and promotion of video game music as an art form."  But does that not extend to EVERY form of music, either human or machine? Why are we segregating music now?

Don't try to throw the OCR mission statement in my face and think that that somehow proves your point. Nowhere in that mission statement does it say that we need to treat all forms of music in the world equally, regardless of whether it was written by a person or generated by a machine-learning algorithm. That's not there. We are dedicated to the appreciation of and promotion of video game music as an artform. This has, for the last 25 years, meant recognizing that the composers who wrote music for video games created art. They are artists. They are human beings who made thoughtful decisions about the music they wrote. They created with intent.

Can you use machine-learning tools to create artwork with intent? Sure. I pretty clearly stated in the announcement that I don't have a problem with machine learning as a technology, but that I don't want people to post music on OCR from services that trained their models on music that was not provided with consent from the artists who created that music. The technology is not the problem; the people building these services using resources they're not supposed to are the problem. These folks have not figured out how to build these services without plundering the collective creative works on the internet. No thanks.