All of the pictures at that link were created by a StyleGAN neural network trained on a lot of art. The point of a StyleGAN is to learn a style and automatically make more works in that style; if you feed it a lot of pictures of people, you get This Person Does Not Exist, which automatically creates pictures of people who do not exist.
It’s entirely abstract art, and does a good job of imitating Cubism, Suprematism, and other styles which make some people snobbish.
It’s not supposed to be “art”, it’s supposed to be a tool you can use to create art. As a (not very creative) example, you can do stuff like take a snapshot with your camera, and tell the computer to re-create it in the style of Van Gogh. Or you can blend and interpolate different styles or elements, generate random samples of anything (maybe the computer will come up with some good ideas), etc.
I see - each time you reload the page it displays a different image. The non-existent persons are so impressive it’s hard to believe they’re not real. Scary.
This Person Does Not Exist has gotten a lot better…it used to have some significant issues with eyes, glasses, and teeth, but I haven’t hit one with the tell there, this time. (Jewelry still looks weird.)
[Edit - I have now hit a couple where the eyes are off, but it’s still a lot better than before.]
… Yeah, it’s still bad with hands. I got one that looked like a cross between Donald Trump and Owen Wilson after a table saw accident not long after making that post.
There absolutely is artistic intent here; software is as valid a medium as paint; a computer is just as good a tool as a paintbrush.
What you’re talking about is found art, not cg art. And even found art is the product of artistic intent as soon as you put a "frame:–i.e artistic context–around it.
There absolutely is artistic intent here; software is as valid a medium as paint; a computer is just as good a tool as a paintbrush.
What you’re talking about is found art, not cg art. And even found art is the product of artistic intent as soon as you put a “frame”–i.e artistic context–around it.
I’m actually pretty impressed. A lot of these “paintings” manage to capture the inexplicable power of good abstract art. Some of them packed a pretty strong emotional punch. Once again I say, as I have said many times, that the relationship between a piece of art and its audience is the only thing that matters; the artist’s relationship to it is just biography.
The intent which exists is a “generalized intent”, in that no specific piece was intended, but the creation of all of the pieces in an aleatoric process was intended. My point is that it’s pointless to analyze each individual work as the deliberate creation of a human mind, because the StyleGAN software isn’t complicated enough to work like that, but you can absolutely comment on the output as a function of the training material.
On a purely technical level, I’m amazed that a StyleGAN can create high-quality fake photographs of human faces which vault over the uncanny valley so comprehensively they’re indistinguishable from real people except in certain instances, but I think creating paintings is more important in terms of what it shows about what makes paintings meaningful and, potentially, how creativity works.
I’m still disagreeing. The concept of the Rohrschach Test is the single most important idea in the entire universe of art. The human mind that creates the emotional content of any conceivable piece of art is the mind of the viewer/auditor/“experiencer” and not that of the artist. Analyzing a work of art is 100% on the audience’s side of the scrim.
I did two dozen “this horse does not exist” and got about one third fairly horselike images, one third uncannily wrong horses or riders and one third very wrong horses.
Like a head at each end, or six legs, or two bums no head wrong. Funny, and shows you how “dumb” artificial intelligence is.