I'm afraid the imbuement we think we feel when listening to music that makes us want to dance is nothing more than the auditory cortex stimulating the motor cortex as they're both in close proximity. I'm a party pooper!

I am aware that we don't have true AI, it's existence is something I've argued against for over thirty years whenever anybody told me about some new breakthrough in AI. It seems though that in the public realm the definition of AI is changing, as definitions will. In a scientific sense we need a specific set of rules that determine that AI has reached self-awareness and a new Turing test to prove it. You mention the Turing test in relation to music, I'm not sure about that but the actual Turing test seems to have raised debate, there are claims that it's been passed by AI.

I don't necessarily doubt that AI could be trained to create unique music with no reference to human music, such a LLM needs to consist of data that 100% excludes all reference to music (but includes all components of it) and the AI needs to be instructed to put these components together in a way that seems to make mathematical sense. Maybe it will 'learn' from human feedback in a strict 'like' or 'didn't like' response format. That would be interesting. What we have now is a kind of Band in a Box on steroids.

Maybe inputting all this human music into AI is getting in the way of it creating something useful to musicians.

 

Bird song?

 

Written on DSEQ Page:
"DSEQ offers various filter types such as cut, peak and shelf filters. The filter sensitivity determines how strongly the signal is recognized by the dynamic equalizers. DSEQ V2 makes it even easier to tame digital harshness thanks to its smart AI function"

 

What did you tell me? I must have missed it.

 

dikker , schnitzel wiener art ist kein wiener schnitzel ....dont be so naive , forgive me .

 

ads up perfect with your avatar

 

HandPan drums too

 

better poll would be : do you think a one man audio effects dev can code a true AI ?

 

Funny I just pulled out a record to throw on, Cheap Trick's Heaven Tonight, one of the best rock albums ever done, guess where it was recorded..

Man Sabbath did so much coke in that building recording Vol 4. TONS of great fucking records have come out of there. That place needs to be deemed a historical building.

 

depends. people could be talking about a million different things when they say ai. i assume this refers to the new generative stuff, so no, but tools like synplant are very fun and useful (often because of their limitations). wanting to replace human talent with a guesstimating machine is silly at best and outright malicious at worst, and ultimately ai is at its most interesting when being really bad at its job anyways.

plus the music industry has always been bogus. pay attention to the underground, folks.

 

I've never talked about resonance though. My criticism was towards you saying it applies to harmony.
Wiki says this about the octave:

Quoting Cooper, Paul (1973). Perspectives in Music Theory: An Historical-Analytical Approach, p. 16. ISBN 0-396-06752-2

The octave phenomenon is a real thing that persists through cultures and borders all around the world, and it might be unique to mammals, idk if you know any contrary scientific evidence, I'd be more than happy if you could present it to me.

So, just to summarize:

My criticism was and still is that you can't use harmony without refering to the human experience.

 

Last edited: Jul 24, 2024 at 4:13 PM

Maybe you can make a sticky post about AI not being AGI because this question will sure pop up a lot in the future

 

i would argue that humans need training data also to learn. is "actual AI" Artificial general intelligence for you?

 

First, I really hate the word AI, because diffussion models are not inteligent at all. It's a hyped word to make buzz and bucks. Once you start to use them and get your head around how they actually work you see that the point you're talking about, a model that creates without input from our part, any original work, is decades if not centuries ahead. Right now they are brute force noise diffussers based on billions of parameters and that's it. You have to guide and curate them to get any originality out of them.

 

You're making an assumption that it is impossible for AI (maybe not now but maybe in the future as it develops) to make music using 'human' music theory and octaves. I'm confused by the idea that octaves are a mammalian phenomenon because it's quite mathematically quantifiable, just like any other harmonic resonance - it's a real thing, not a human experience. AI might have no way of 'understanding' the relationship of notes from a human perspective if it has no reference to human input/models but that doesn't mean it can't discover this for itself, perhaps even in the same way that we have.

If AI creates a mess of music that is harmonically correct but a jumble of tones - including octaves, then learns from human input of opinion whether humans like each piece or not, it will very quickly determine what works and what doesn't.

The music theorists are out of their depth when it comes to AI, their primary concern is how we like and create music and why it works from a theoritcal viewpoint. The mistake is thinking that only humans can do this. Music is data.

 

But AGI is still AI so within the context of my assertion it's very relevant. More so.