But when i experiment with this AI generated art,i remember a passage in George Orwell 1984 that kind of scare me:
Quote:
“The tune had been haunting London for weeks past. It was one of countless similar songs published for the benefit of the proles by a sub-section of the Music Department. The words of these songs were composed without any human intervention whatever on an instrument known as a versificator. But the woman sang so tunefully as to turn the dreadful rubbish into an almost pleasant sound.”
You read my mind, Hornet, I was just about to look up that passage from Orwell. I have to say this is an interesting experiment in image-generation, but whether the result is “art”, well.. I shudder at the thought of a potentially infinite proliferation of vacuous computer-generated images swamping every eye and brain on the planet. I’m not giving up paper and ink any time soon!
I guess for composing elevator music or creating certain types of stock imagery some of these models are going to get there.
But as many new technologies before this won't "kill art".
The central point being that these systems have no actual representation of the language they use ... it is basically string input. At no point do they "understand" anything in terms of reference, categories etc.
Which leads them to fail hard on simple tasks, if you prompt them with something that is more of a semantic instruction to build an image, than a "vibe" or atmospheric thing.
For instance check out the result for the description
two red spheres placed on top of a larger blue cube, which is resting in a green bowl on a table.
The AI basically picks out some of the shapes and colors but obviously has no internal world-model to represent the scene at all. (it doesn't even really know what follows from something being a cube...)
Now try to give it instructions to build just a simple cross, by describing only the dimensions of the wooden beams, and how they are joined together (angle, and place). Without mentioning the word 'cross'...
They're good at creepy moods though.
One approach to make use of them creatively is to selectively use their output as collage components.
If you let them create 'eldritch horrors' sometimes these come out in a diffuse disturbing way that's very useful for that genre. One might use that as a background or element of a larger composition.
Another artistic aspect is, in awareness of their basic stupidity, trying to coax them into creating things that are otherwise hard to get out of them by manipulating the prompts.
Same goes for the GPT models, they're basically bullshit artists.
(Which means they can write articles that sound reasonably convincingly like boilerplate political reporting)
Dr Who as a Batman vilain)
