Artificial Intelligence and music creativity

Post Reply New Topic

Carl's Jr made it about 6-7 years ago. Tastes like McDonalds to me, but their shakes and fries are out of this world.

no carls in my neck of the woods or anywhere near. but our local mickey-d's have half-price deals all the time on anything they sell, buy one and get one free.

I have been trying to replicate the Micky-D's special sauce but to no avail

https://colab.research.google.com/github/alembics/disco-diffusion/blob/main/Disco_Diffusion.ipynb#scrollTo=DiffuseTop


Kinkade is actually the default string for the text prompt.
I've been playing with this yesterday and today. it takes a while, and it seems to be struggling with sort of the overall concepts of how things are connected - but you can actually put in a sketch as initial image, and just leave the detailing to discodiffusion. it's kind of a fractal thing, though, adding ever smaller detail until you stop it (or rather, until it reaches the maxsteps value)

yeah... so... Van Gogh is extremely decorative, if it weren't for his life's story. Ai won't be able to actually live a life, but it can easily invent an artist's life story to go with the paintings (and quite frankly: I have only read about Van Gogh. For all I care, he could have never existed etc. - like Shakespeare, maybe, who knows)....
But: that means the art isn't actually in the paintings, it's in the life of the artist.
An AI can thus only produce images, but not art.

that's actually where art has gone since Andy Warhol: to be about the artist and the image he creates of himself, rather than the artworks he creates.

Beauty/art is in the eye of the beholder. There is an authentic element when the mind translates onto canvas. That is what the buyer is paying for when they buy art.

Art theory considers "authenticity" an effect that can be created at will - and artists often consciously do. - but if there was something to the concept of authenticity, then it wouldn't be in the eye of the beholder, right?

But beauty isn't solely in the eye of the beholder, there's also an environment that imprinted on the beholder to shape his taste, which includes the taste for authenticity, whether that exists or not becomes irrelevant, if it is only one of the subjective criteria applied in individual judgement.

Yes you are correct...but, artwork (whether canvass, music, writing or sculpture) is still subject to the marketplace. There will be people who will pay big bucks for reputable artist created work Vs mass produced computer generated art,

Nobody is going to bid at an art auction for something that can be mass produced.

Yes, .... Except for bored ape NFTs, but the NFT market has collapsed recently, even before the current crypto-crash.

But that's what I meant when I wrote that the artwork itself is kinda irrelevant today, and it's all about the artist's image/reputation, price speculation in a market, and actually also money laundering. Also art-washing (green-washing, just with art endowments. The Sackler family sponsored quite a few art institutions with their oxycontin-drugmoney).

Here's some examples of what Psylocybin can induce in a shamanic artist painting the spirit world existing in parallel with the real world of his south American home (ironically this concept was copied by the writers of the Netflix hit series Stranger Things).







Computers will not be generate hallucinogenic induced artwork because we don't know how drugs induce the brain to create these images (or even if this is a real place?) so you can't program AI to make this stuff up.

What do you mean? - those images are merely bunching together things, somewhat randomly, and the artist claims to have been inspired by drugs. But if I didn't know that, and the artist had said he had rolled a dice to determine what to paint next, that might be just as valid an explanation.
Salvador Dalí has claimed to be able to think like a schizophrenic without actually being schizophrenic, and that's how he got inspired.

Anything I type into disco diffusion comes out looking like the pictures you posted, plus some Dalí-esque distortions of shapes.
To me these images look like the very counter-example, because they are exactly what AI *can* do right now. Except, AI can add the fine structure of a renaissance freso, watercolour, oil or digital painting, or make it look like a photograph, if you ask for it.

Funnily, what it struggles with are very basic concepts: make an image of a red cube on top of a blue cube - it will give you red and blue cubes, in varying positions and numbers. Sometimes it will be the red on top of the blue, but nowhere near always.
That might however be a problem of the language processing and it's connection to the image processing. My guess is next generation will combine the generative capabilities of clip guided diffusion like Dalle-2 with the self-checking mechanisms of GANs....


Edit: the art/design youtubers are discussing how this could aid the early design process. Which... You know... That's where the creativity happens. After the early phase comes the tuning, the figuring out how to increase effects or how to make the idea fit the message/intended meaning.
But while they're talking how this can *never* replace human creativity, they are literally discussing how this can be used to replace the idea-finding process in art-creation.

Yes it's possible the artist is saying it was induced by mushrooms. But then again he is painting what he saw in the alternate reality/underworld (In Stranger Things they call it the "upside down").

Humans aren't the only organisms to make inspiring art under the influence. Here's a creation by an orb weaving spider under the influence of various drugs

Ah. I see. You seem to be believing in drugs being a gateway to alternate realities, where I see them as disturbing signal processing in the senses and brain. As in: there's no alternate reality, you're just rewiring the brain for a bit to see what happens.
If you rub your eyeballs until geometric patterns emerge, that's because you're stimulating the ganglia in your retina, but without a particular signal, so all the neurons are firing - ganglia in the retina are interconnected and function as image sharpening kernels, they block or enhance neighbouring inputs- so when you press on them, you're pushing all the sharpening effects at once, on random noise coming from rods and cones, resulting in these geometric patterns that are just forming like waves.
It has been noted that visually, it resembles DMT-induced hallucinations quite a lot, except the hallucinations are a lot more colourful.
... I don't see why

Same for spiders, but It should also be mentioned: spiders don't make webs for artistic purposes, but to hunt for food. Their shape is algorithmically determined.
Throw drugs in there and it just adds some random vualues into the "if/else" loop, and the result will be... Well, different.

If you look at creativity there seems to be a chemical or psychological determinant. I used to hang out with an artist and he used various stimulants to paint. Often his best work came when he was in a particular "mood", That mood could be enhanced chemically or from personal experiences he had.

Artists like Goya painted a particular darkness that came from his own experiences, not from simple imagination or execution of some artistry/technical skill. And of course we know what inspired Van Gough or Edvard Munch.
So music is the same. The late rapper XXXtentacion regularly experienced mood swings and suicidal ideation and this came out in his music which is popular among young people experiencing similar levels of depression.

The idea that somebody could have an authentic emotional experience from a computer generated song or painting is unlikely BTW the spiders webs are functional but the patterns might involve some imagination. Like an elephant who paints.

I fed a depth-image of a face into disco diffusion and asked it to make a michelangelo from it. The depth image was white-near and black-far, so everything but the face was black. What came out looked like Francis bacon crossed with a Michelangelo marble statue- a twisted and torn face, with alabaster skin. It didn't have the energy of a Bacon painting, but it could easily serve to freak someone with less tolerance out. The fade into the pitch black background made it quite haunting.

I really don't understand how people can not see that AI is creating these at random, but if you ask it to do a hundred variations, randomly, it will generate something good. Infinite monkeys on typewriters, basically, trained on 400 million images.... For now. A dataset with 4 billion images rxists, but no one has trained his algorithms to completion on that yet.