I think the computers just won.

[auxien]

21 hours ago, logakght said:

What AI will never understand is the "symbolic" and "imaginary" dimensions of art creation. It is as real as you and me creating music, but it lives in another realm of "enjoyment".

the interesting thing about that is that we often don't understand our own symbolic/imaginary dimensions of art creation. a simple metaphor? sure, yeah, i get that, a poem about sunsets could also be about a death, or a relationship that's over, or whatever symbolism interpreted by the reader. Chat GPT and such already understands that symbolism is very much a part of most art/etc....it's not yet conscious tho, so it doesn't 'get' it, it just knows that it is a thing. but my point there is that often, neither do we. even the creators of the art. so your point maybe has something to it, or maybe not...

but once these creations are complex enough to become self aware/conscious... well, all that's out the window. a computer program's 'enjoyment' could be vastly different than ours once it does have some analogue of 'enjoyment' ...or it could be very similar to human concepts (it will have been trained almost entirely on human-created input of course) or similar in the ways that many animal's enjoyment is similar to our own. the body thing is pretty key to this concept tho, i'm still not sure consciousness will ever be possible without live autonomous sensory inputs, a true physical 'self'.

21 hours ago, taphead said:

I'm impatient and didn't watch that video but did Neely mention George Lewis' Voyager stuff in this context because it seems relevant

no i don't think so

 

[Enthusiast]

 

[BlockUser]

On 4/15/2024 at 9:32 PM, Brisbot said:

What is needed from Ai is a much quicker way to get your ideas down, and to realize them FULLY. Would love to be able to give it a short one shot, then with a prompt as it to generate tons of variations. I mean I already kinda do that when you throw FX on existing sounds, record it, resample it, etc. I guess it would just be quicker. If I have a specific idea in mind then any tool can only bring that idea about quicker.

But why is this needed? The way you're talking about ideas and getting them down, it sounds like you see art as a product. The way I see it, art isn't a product, or an object. It's a process. Ideas change a lot when you're really down in the process. Instinct, mistakes, resistances (inner and outer): All these are parts of the process itself, they aren't hindrances. More often than not, you can incorporate these. A synth isn't a set of limitations that keep you away from making music. It's a set of definitions and pathways that enable you to make music. Yes, it takes time and humility. Yes, it can do your head in, especially if your ego is getting in the way, and I'd say that ideas hindering the process is definitely the ego sabotaging the process. There is no art without the process. Sure you're talking about art and not design?

[BlockUser]

On 5/2/2024 at 1:12 AM, auxien said:

the interesting thing about that is that we often don't understand our own symbolic/imaginary dimensions of art creation. 

but my point there is that often, neither do we. even the creators of the art.

the body thing is pretty key to this concept tho, i'm still not sure consciousness will ever be possible without live autonomous sensory inputs, a true physical 'self'.

 

It really is interesting. But as humans, we don't need to understand the symbolic / imaginary dimensions in order to create. Creation and analysis are 2 different things. AI relies on imitation and pastiche. When it generates an image, it only works on the basis of the image, not on the basis of what contributed to the making of the image.

Totally agree with the body thing.

[hoggy]

I mean, notice how all the music sucks though

But seriously check out Adam Neely's latest vid - I think he makes a great case for why there's some things AI (which is not AI, it's machine learning only) will not be able to do - I especially like the section about 4 types of cognition
 

Oh, I see auxien already posted it - but yeah, the concept of "musicking" vs. music as a product/output is important

Edited May 4 by hoggy

[auxien]

1 hour ago, BlockUser said:

AI relies on imitation and pastiche. When it generates an image, it only works on the basis of the image, not on the basis of what contributed to the making of the image.

for now, yeah. a few more layers of complexity/abstraction/'understanding' and it'll be indistinguishable from a human's creation.

those 'layers of complexity' are going to be exponentially harder to get to, of course...but if that's not already happened in a basement somewhere, it's very likely it could happen in our lifetimes.

34 minutes ago, hoggy said:

But seriously check out Adam Neely's latest vid - I think he makes a great case for why there's some things AI (which is not AI, it's machine learning only) will not be able to do - I especially like the section about 4 types of cognition

that's what we're talking about, scroll back to the last page my argument is that Neeley's talking about where things are at now, and can't (or doesn't want to) see past that to the likely future of the tech. i like Neeley's points in the vid tho, he's not wrong at all, just a bit shortsighted perhaps.

[hoggy]

18 minutes ago, auxien said:

that's what we're talking about, scroll back to the last page my argument is that Neeley's talking about where things are at now, and can't (or doesn't want to) see past that to the likely future of the tech. i like Neeley's points in the vid tho, he's not wrong at all, just a bit shortsighted perhaps.

I mean if they make androids with feelings that have a subjective relationship with the world (we're WAY off from that), maybe they will be able to make something meaningful - but for now machine learning generated stuff is only a tool - because of all the interaction that's needed for live performance or social connection around music - like I'm sure there'll be AI generated songs I like at some point in the future, but like Adam says, it's not all about the output/product - it's not all about recorded music - music is something people participate in, and for now, ML stuff is not an active participant

to be clear, I don't totally agree with Adam, I don't think machines will never ever pass the musical Turing test (the output test soon sure, although making an evolving body of work showing consistent depth and emotional development, no), but I think it will be hundreds of years from now before true AGI

Edited May 4 by hoggy

[Rubin Farr]

Skinny Puppy consider whether or not to get an adult circumcision

https://www.udio.com/songs/1wxDdSRVyHfg4uGfaVpQL2

[Rubin Farr]

Kristi Noem feels really bad about killing her dog, but he was woke, so he had to go.

https://www.udio.com/songs/7yu3kW8iuDpt5JRjkDsmeL

[auxien]

slop, i love it.

[Psychotronic]

Oh, so this tool just enabled humans to memifiy music, as everything else that is real on the Internet is basically a meme.

Onwards!!

[jaderpansen]

paper: https://arxiv.org/abs/2404.04125

would be quite hilarous tbh.

[chenGOD]

So if I understand the video and paper correctly, Generative AI is starting to hit diminishing returns through current learning methods?

[EdamAnchorman]

1 hour ago, chenGOD said:

So if I understand the video and paper correctly, Generative AI is starting to hit diminishing returns through current learning methods?

My totally naive reading of the abstract and conclusion is that current models cannot accurately generate images from text-based concepts that it hasn't explicitly seen before, and to get it to do that well we would need exponentially larger training datasets.

edit:  didn't watch the video; have no idea what that's about

Edited May 13 by EdamAnchorman

[Rubin Farr]

https://stability.ai/news/introducing-stable-audio-open

[chenGOD]

14 hours ago, Rubin Farr said:

https://stability.ai/news/introducing-stable-audio-open

See now this is maybe an actually interesting use of AI for music. I wonder if its synthesis engine can morph sounds.

[Rubin Farr]