Live data from Hacker News

A Mathematical Theory for Why People Hallucinate

quantamagazine.org

1–10 of 46 posts

Re: A Mathematical Theory for Why People Hallucinate

#2
If you look at how neural nets work, they too tend to hallucinate. It doesn't seem so mysterious to me. We have GANs, autoencoders and network visualisation tools (finding the input that maximises a certain neuron or set of neurons) - and they show that noise injection into the net can cause hallucinations.

Re: A Mathematical Theory for Why People Hallucinate

#3
I've held this theory on my own after experiencing geometric shapes while hallucinating after smoking a lot of marijuana years back. I sat down closed my eyes and experienced an intense feeling of flying through what I can only explain as the starfield screen saver from Windows but with colored 2D shapes heavily dominated by hexagons of various sizes. During that experience I realized that I might be experiencing the inner workings of my brain in real time. I didn't give it much more thought but that ideas stuck in my mind for years. Recently I began fitness training with a trainer who believes in the stoned ape theory which reminded me of my own experience and I again read into it a bit. I'm not sold on the idea but the overall theory is solid in my book. Still to this day if I get really high I experience geometric hallucinations.

Re: A Mathematical Theory for Why People Hallucinate

#4
> The sparseness of connections between inhibitory neurons prevents inhibitory signals from traveling long distances, disrupting the stochastic Turing mechanism and the perception of funnels, cobwebs, spirals and so forth. The dominant patterns that spread through the network will be based on external stimuli — a very good thing for survival, since you want to be able to spot a snake and not be distracted by a pretty spiral shape.

I see spirals every time I meditate. I find it interesting the suggestion that meditation allows more random noise into my neurons.

It feels very similar to the dream state where the content usually has clear connections to my real life but is randomized just enough to produce some odd effects.

In the end, understanding dreams or hallucinations feels like interpreting art to me.

Re: A Mathematical Theory for Why People Hallucinate

#7

I've held this theory on my own after experiencing geometric shapes while hallucinating after smoking a lot of marijuana years back. I sat down closed my eyes and experienced an intense feeling of flying through what I can only explain as the starfield screen saver from Windows but with colored 2D shapes heavily dominated by hexagons of various sizes. During that experience I realized that I might be experiencing the…

I had a similar experience when I forgot to take my SNRI antidepressant. I woke up in the middle of the night, and when I tried to continue sleeping I had these ultra-detailed geometric patterns dancing before me. The visuals were nice, but then my ears started ringing too which started bothering me, so I opened my eyes and woke up completely, which stopped the visuals and the ringing.

Re: A Mathematical Theory for Why People Hallucinate

#9
Is this 'hallucination' though? Or just some weird stuff happening in visual processing?

I underwent some very serious periods of sleep deprivation (Army) and our 'hallucinations' were of a totally different kind: having conversations with people that were not there, seeing things that were not there, misidentifying people whom you know really well who are right on front of your face, taking on different personalities etc..

Re: A Mathematical Theory for Why People Hallucinate

#10
I wonder if the same effect can be reproduced with image recognition software.

The "Google deep dream" look very... acidic to me, but there were deliberately induced by reinforcing a particular signal. But I wouldn't be surprised to get similar effects by adding noise to the proper type of neural network.

Post reply on HN