Google’s hallucinating neural networks suggest psychedelic experiences are spooky but not supernatural.

I have always believed psychedelic visions from ayahuasca etc. are not supernatural. Google just proved it.

Computer image recognition has transformed out of all recognition in the last few years thanks to deep neural networks and massive sets of training data.

We train an artificial neural network by showing it millions of training examples and gradually adjusting the network parameters until it gives the classifications we want. The network typically consists of 10-30 stacked layers of artificial neurons. Each image is fed into the input layer, which then talks to the next layer, until eventually the “output” layer is reached. The network’s “answer”

Instead of exactly prescribing which feature we want the network to amplify, we can also let the network make that decision. In this case we simply feed the network an arbitrary image or photo and let the network analyze the picture. We then pick a layer and ask the network to enhance whatever it detected. Each layer of the network deals with features at a different level of abstraction, so the complexity of features we generate depends on which layer we choose to enhance.

One way to visualize what goes on is to turn the network upside down and ask it to enhance an input image in such a way as to elicit a particular interpretation. Say you want to know what sort of image would result in “Banana.” Start with an image full of random noise, then gradually tweak the image towards what the neural net considers a banana

The results are intriguing—even a relatively simple neural network can be used to over-interpret an image, just like as children we enjoyed watching clouds and interpreting the random shapes. This network was trained mostly on images of animals, so naturally it tends to interpret shapes as animals

Inceptionism: Going Deeper into Neural Networks

All very trippy, I think you’ll agree.  I consider this strong circumstantial evidence against all the supernatural woo associated with ayahuasca etc. If computers hallucinate as awesomely as this. it suggests the human brain doesn’t need any external help generating the self-transforming machine elves and other crazy visions often attributed to aliens or inner worlds.

Source: Photos by Michael Tyka – Google Photos

About caspar

Caspar is just one monkey among billions. Battering his keyboard without expectations even of peanuts, let alone of aping the Immortal Bard. By day he is an infantologist at Birkbeck Babylab, by night he runs
This entry was posted in good. Bookmark the permalink.

3 Responses to Google’s hallucinating neural networks suggest psychedelic experiences are spooky but not supernatural.

  1. steeplemouth1 says:

    Interesting. Those images are indeed trippy and vaguely reminiscent of visual disturbances created by a low dose of a psychedelic compound, but it is hardly proof that hallucinogenic visions are simply elaborate pareidolic images constructed from random elements in the visual field. What about closed-eye hallucinations, which tend to be the most baroque and complex hallucinations? How can the brain built those astonishingly beautiful, incredibly rich and detailed images in the absence of any visual stimuli whatsoever? I’d be interested to hear if the author of this article has ever had a high-dose psychedelic experience. I’m willing to bet that the answer is no. The difference between a visual disturbance resulting from a low dose of a psychedelic compound (such as the effects in the image above) compared to a full-blown hallucination resulting from a high dose is like the difference between a child’s drawing of a stick figure compared to a Carravagio or a da Vinci oil painting.


  2. dondeg says:

    Don’t forget, its your human mind that is interpreting the image! To the computer, it’s just patterns of 1s and 0s. I think attributing hallucinations to computers is much more far out that assuming that psychedelic visions are of other worlds.

  3. Thanks for your comment, Steeplemouth.
    Sorry for delay in responding, I’ve had wordpress problems.

    As it happens I have had plenty of high dose experiences and I stand by the original claim. The top two images in this post look a bit the ‘walls are melting’, ‘faces in the clouds’ experience you might get overlaying your visual experience during an eyes open hallucination. But the other two images on this page (and many, many more of the images produced by this approach, search for the hashtag #deepdream) are nothing like that they are much richer and stranger and yet are still just a product of the network trying to make sense of a fairly senseless input. Often in fact created by the network reinterpreting it’s own outputs much like a dream or closed eyed psychedelic experience.) So I’d say that they correspond very well to the baroque complexity of a high dose closed eye experience. And remember that all of these incredibly complex yet utterly alien images were created by a network that has only had experience of natural images. there’s nothing spooky in the input but the output is decidedly strange.

    True that this is hardly ‘proof’, but to me it is very suggestive and simple explanation of machine elves etc. (And I think William of Occam would agree.)

    As another example see what the network dreams as it closes its eyes and zooms in from a single image of me and a friend 🙂

Leave a Reply