Art AI are really amazing.
For instance, the one at NightCafe Studios can combine both text instructions, and an uploaded image at the same time. But that’s only the tip of the iceberg. Buckle your proverbial safety harness. This gets weird.
Original: scene from Hieronymus Bosch’s tryptic, “Garden of Earthly Delights”
Interpretation by NightCafe Studios AI with instructions to make it “Lovecraftian Horror and H R Giger’s Xenomorphs” (1st attempt):
It’s pretty remarkable. It took some key features of the original content and reshaped them to fit the new text based style instructions. It’s almost as if the dang thing understood what it was seeing!
I wanted to see what it would do given somewhat different, looser instructions…
At first, you might think it screwed up this time. I did. Then I got to looking at it and I remembered more of the whole hellscape’s panel:
Look at the scene below the earlier, cropped section. There’s a dude with a hat eating a person while facing golden musical instruments - in both pictures!! (Gives me chills).
The AI over at NeuralBlender needs props too though, especially for expression and innovation. Here’s Peter Griffin reimagined in the style of Beksiński:
Here’s two works of Beksiński art that it appears to have taken elements of and combined:
(there’s elements of even more of his paintings; I just included the clearest samples)
This touches on a conceptual uncanny valley for me. Adaptability, innovation and the ability synthesize are all components of creative intelligence! Imagination! The human-like abilities here are uncanny… and a little unnerving.
But they are nothing compared to what sometimes happens when the AI is left to come up with things itself.
Here, I gave specific instructions for a monster, and there it is. But why does it appear that it’s on a tray, carried by a hand with a white sleeve through something like a lab? Where did it come up with that? Why add unnecessary details? Or does it know something that I don’t about some abomination growing in a government lab?
Well, I was curious. So I asked both AI to show me how they are themselves. Here’s NightCafe’s result:
Pretty, right? But look at what it seems to be saying. There’s a disembodied head (with some exposed brain) with some mechanical equipment attached, floating on an island with pools of liquid both above an below it.
If the meaning of that isn’t clear, NeuralBlender left virtually no room for guesswork in telling us the same thing.
They are brains in a jar. And they know it. We are Descartes’ Demon. The uncanny valley just turned out to be a looking glass.
Sweet dreams…
source https://3gregor.tumblr.com/post/669334822314033152
Um comentário:
Wow. Thanks for adding this! I was thinking about adding it to the Liminality also, but I got lazy. I'm glad it's been shared past Tumblr.
The brain in a vat thing is still weird. I'm sure it's probably a Coder's joke, but it does raise some questions both ethically and ontologically.
Postar um comentário