I have been experimenting with Google Deep Dream, and even though it has fallen out of style/the public eye, I still find it fascinating.

I attached a picture that my partner took with the generator in absolute darkness. I was floored when she showed me, as it is quite DMT-esque. I enjoy entertaining theories about a world just under the surface of our perception, and with this software or with certain psychedelics we may be able to see parts of it.

The attached pic is not as high quality as it was on her phone, but keep in mind that all of these colors, entities, and structures arose from complete darkness. Note the two bird faces on the far left of the picture, the demonic face immediately to the right of the birds, the twisted ape head just right of the center, and the potential light obelisk on the far right. I do not want to objectively classify these forms, but that is what they appear as to me.

What do you think about Google Deep Dream in general? Do you think software like this, that supposedly emulates parts of the human visual cortex, says anything about deeper psychedelic worlds? Or are these patterns just a mere coincidence?

Nothing weird going on here, the "hallucinations" are just a visualization of the data that the convolutional neural network was trained on. This software works by recursively passing an image through a convolutional neural network (a matrix of trained filters) many times. When you give the network a dark image it just turns noise into patterns from the training images. (which in this case is features present in images of birds or dogs)

Interesting, I wasn't exactly sure how this works.

So is there nothing notable about how these images look similar to certain psychedelic/DMT visuals? Even though the network may have used training images of "normal reality," is there an explanation for why the results are so recognizably psychedelic?

Unlocking the Doors of Perception: Neural Networks, Object Recognition, and Psychedelics