Tron man right center.
Also this is what reality really looks like without the narrow meat filter you call human perception.
4 Likes
Ah, yes – the universe is not only stranger than you think, but stranger than you can think…
4 Likes
That’s as well as may be but nothing, nothing, has ever tripped me out as much as this one:
DMT, a heroic dose, whatever, stare at your slug-dogs and giggle for a minute, then lose your shit at the milky way. If you can escape the synaesthesic hum of this picture then you are a faster person than I.
3 Likes
I saw that image tweeted recently, with a note that it was generated entirely by computer. I mean, okay, so was the near-photo realistic picture of a medieval village in a game I was playing the other day. I have no idea what the story is with this image.
What were they actually trying to accomplish here? Was this a success or a failure?
It looks happy and relaxed. Maybe needs a cigarette.
Looks to me like it’s a neural network trying desperately to look for human faces, and mostly outputting Pareidolia.
I like to think that this is the way Alex Jones must see the world.
5 Likes
My guess was that it’s some algorithm intended to recognize distinct objects in a set of photographs, and generate a new image with those objects in it. There are a few bits that look somewhat like a table or kitchen counter and the objects upon it, and then there’s what looks like a squirrel that made direct eye contact with Cthulhu.
3 Likes
Fractal slug puppies
2 Likes
2 Likes
1 Like
If I were asked to design software to do this, it would use trained neural nets to focus on part of an existing image and find the best fit, then subtract the error signal from the neural net from the image in that spot. The chosen NN could be applied to the image like a brush, fitting and subtracting error from whatever region the user is painting. Switching to different neural networks that have been trained to recognize different features would allow the artist to paint eyes in one area and leaves (or whatever) in another.
While the image bears some resemblance to the generative images in the paper, I have yet to see the image within the paper – v1 or v2.
The reddit page sez:
Samples from this paper look similar, but not as detailed and intricate as the multi-scale dog-slug posted on imgur. Any idea where the difference lie? Longer / better convergence? Larger models?
Also the resolution is much higher than in the paper.
Actually, much of the comments seem to be debating the origin of the image.
Coincidentally (OR CONSPIRACYICALLY) I had the paper and image open in my browser since last week…
1 Like
It’s what craft services came up with to feed John Carpenter’s dog-thing. (And please, trust me on this – you do not wanna see what they served Wilford Brimley.)
2 Likes
Just saw this in my tumblr dash:
Google Research release images related to their work in Neural Networks - just as they are used for image recognition and learning, when they are used for image creation, the results are incredibly surreal:
More at the Google Research blog here
7 Likes
You beat me me to it, Nelsie! Still haven’t found the original image, but I think we’ve got the right ballpark.
1 Like
Reminds me of Barnsley’s old Fractal image-compression scheme from back in the dark ages of the 1990s, when the idea was to compress an image by representing it as a combination of small, distorted copies of another image.
1 Like
1 Like
