Slashdot Mirror
Turning Neural Networks Upside Down Produces Psychedelic Visuals
cjellibebi writes: Neural networks that were designed to recognize images also hold some interesting capabilities for generating them. If you run them backwards, they turn out to be capable of enhancing existing images to resemble the images they were meant to try and recognize. The results are pretty trippy. A Google Research blog post explains the research in great detail. There are pictures, and even a video. The Guardian has a digested article for the less tech-savvy.
I've had a few up-close experiences with heavy psychedelics. Those photos took me right back. Wonderful insights!
I ran slashdot backwards through the DICE marketing bullshit neural network and got www.soylentnews.org
The slashdot UI produces psychedelic visuals even without any artificial or natural intelligence.
Every time I come here, there are icons all over the place, in the middle of the text, the title bar shows random icons or text and I'm not even on beta.
Not to mention the dupes or stupid articles and don't make me begin about the videos.
A human can't do it? Alex Gray begs to differ.
;-).
I guess we could argue that it's "similar" (i.e. not the same), but it's pretty darn close
The Mandelbrot set is a very different animal from what these algorithms are doing. I agree that a human couldn't draw a Mandlebrot set, but in some sense this work is much less precise and analytic than something like a Mandlebrot set.
This makes me wonder if a similar process is occurring in the brain of someone on a psychedelic. Are the compounds stimulating pattern recognition feedback loops from the inside out, causing people to see their imaginations manifested in the fuzz?
Any possibility that they will release higher-res versions of these images? Maybe sell some prints?
I realize these are just the output of a funnel-run-backwards, but they'd make awfully cool posters.
An internal system operation returned the error "The operation completed successfully.".
I have seen things very similar to these on some level. My pattern-recognition algorithms aren't fixated on dogs, but the way some of these images look is very familiar. Translucent objects popping out of feedback-looped noise is real. Fractal-repeated infinite patterns are real. Sky turning into crinkly paint swirls is real. Empty space around objects (like the Google logo) showing iridescent patterns is real. Now, there is a caveat that each experience I've had was as different from the previous one as it was from the default world. I also know people whose experiences are vastly different from mine, or even completely non-visual. Not all brains are wired the same!
To some extent yes, but it's likely way more complicated than that. But, yeah, without sensory input we start hallucinating. It's like asking your senses repeatedly "is that real" and the senses always say yes, rather than no. So you drift off into whatever because that's real and therefore there's this other things too.
It's a bit like that old parlor game where you tell somebody that you're going to have them ask questions about a dream, send them in the other room, asking for dream volunteers, and then tell the people still in the room that the answer is yes if the last letter of the last word of the question ends A-M and no if the last letter of the last word of the question ends N-Z. -- They inevitably guess a dream involving all manner of perverted stuff as the crowd confirms and rejects bits at random. Inventing a dream out of his own head rather than somebody else's head.
There's also every day hallucinations like seeing detail where it doesn't exist, movement where it doesn't exist, and hallucinating something to fill the big blind spot in our eyes.
It is no longer uncommon to be uncommon.
Louis Wain:
There has been some speculation that Wain’s schizophrenia was caused by toxoplasma gondii—a parasite found in cat’s excreta. Whatever began the illness, Wain was incarcerated in various asylums and mental hospitals for years at a time. The changes to his life were reflected in his art. His paintings of cats took on a radiance and vitality never before seen: the fur sharp and colorful, the eyes brilliant, and a wired sense of unease of disaster about to unfold.
But these paintings look normal compared to the psychedelic fractals and spirals that followed. Though these are beautiful images, startling, stunning, shocking—they suggest a mind that has broken reality down to its atomic level.
Though it is believed that Louis Wain’s paintings followed a direct line towards schizophrenia, it is actually not known in which order Wain painted his pictures. Like his finances, Wain’s mental state was erratic throughout his life, which may explain the changes back and forth between cute and cuddly and abstract and psychedelic. No matter, the are beautiful, kaleidoscopic, disturbing and utterly mesmerizing.
I know nothing about these NNs, but the NNs used for the ImageNet Competition typically have a few hundred thousand neurons. This is to place images of about 1M pixels into one of 1000 categories. Most image recognition NNs are "convolutional" which means they are tiled. So each neuron in the first layer is only looking at a small part of the image. Later layers will pool the results from the convolutional layer into extended features. This cuts way down on both the size, and the amount of computation needed. The number of layers will typically be 5 or 6. Even more layers should, in theory, help, but deeper networks are very hard to train. The total size in bytes would be maybe 100MB, but that will vary widely depending on the implementation. I don't know how big Baidu's implementation was (they were the winner, but they cheated). NNs can run fast, categorizing hundreds or thousands of images per second. But it can take a long time to train them, days or weeks on a GPU farm. Fortunately the training is highly parallel.
Does anyone know if these images can be created in real-time? If so, demo-coders will pounce on the algorithm and have an absolute field-day! Demos will never quite be the same again. Another idea could be an easter-egg for a video-game where if the player has just ended a very intense gaming-session, the visuals of the frontend (even if only the background) could have this algorithm applied to them just to see if the player notices anything out of the ordinary (after a particularly intense session, this will be harder to spot immediately).
I know that training a neural network can take a very long time, but using it to recognise images can be done very quickly. If a standard CPU or GPU cannot do this in realtime, would the more dedicated demo-coders start creating their own FGASs / ASICs that are designed just for this task, and bringing their creations along to demoparties?
Perhaps the input images they used were also low-res? If they had used higher resolution photos it would have taken much more computing time to run them through the neural network for hundreds of iterations. I guess the same neural networks could also enhance the resolution of the images by being fed a scaled-up version and outputting it with more (imagined) detail.
-- Ed Avis ed@membled.com