|
|
|
|
|
|
by GuB-42
511 days ago
|
|
|
Maybe the most "scientifically accurate" replication of psychedelics are in these "DeepDream" images. They were originally made to debug neural networks for image recognition. The idea is run the neural network in reverse while amplifying certain aspects, to get an idea on what it "sees". So if you are trying to recognize dogs, running the network in reverse will increase the "dogginess" of the image, revealing an image full of dog features. Depending on the layer on which you work, you may get some very recognizable dog faces, or something more abstract. The result is very psychedelic. It may not be the most faithful representation of an acid trip, but it is close. The interesting part is that it wasn't intended to simulate an acid trip. The neural network is loosely modeled after human vision, and messing with the artificial neurons have an effect similar to how some drugs mess with our natural neurons. |
|
|