- They're already well able to surface the most addictive short video for a specific user out of millions of real videos.

- But these millions of real videos are just darts thrown into the space of "videos that could hook the user", in the end even the best-selected of them is not perfect.

- Now, behold! AI allows to generate the perfect video to surgically hit all the switches in the viewer's brain and turn it into a zombie hooked for days on end.

Let's hope our regulations hit these "social networks" hard enough so that never dare deploy this kind of technology.

We train an encoding model, a “digital twin”, that predicts how each visual region responds to any video. Now we can ask: which video would make a chosen region light up the most? NEvo searches for that video automatically, using the twin’s prediction as its reward.

I only scanned the paper, so maybe I missed it, but is there any confirmation that this ‘digital twin’ works? Like, do the generated videos actually cause the same patterns as in the ‘digital twin’ brain model in real humans in an MRI machine? My instinct is to be skeptical that it’s possible to reliably create a video -> brain activation prediction model.

This is a tool to help researchers in figuring out what different parts of the brain are actually for with less experimenter bias contamination of “well we think maybe it’s about this so let’s show it video of x to see”.

The essence runs on having someone sit in a scanner for a couple hours watching all sorts of things, and then feeding that to a model that will then build its own representation of said data and try different things on it until it’s found what makes a certain part sing in the model.

The purpose is a generalized understanding of brain function, more or less the same way we’ve been doing it all these years. Expose brain to something, record it somehow, see if brains reaction in the recording helps you understand more about who we are and what cognition is.

I had an Aunt who had dementia. This is obviously a terrible outlook. But her and my Uncle seemed to be doing okay. My Uncle was an ultra-competent guy, highly stable, the person you could rely on; and had been his whole adult life. So it was shocking when he had a mental breakdown and became manic. There was probably something physical going on, but what was also going on was that he had wanted my aunt to be able to live normally for as long as possible, so had been covering everything - for a year or more he'd had to be alert 7 days a week in case my aunt tried to cook on the gas stove or something like that, at which she was no longer safe. So the risk-alert part of his brain had been constantly overworked.

I appreciate that this is scientific research, but there are definitely companies out there that will try to row-hammer everyone's brain if this sort of thing is not heavily controlled.

https://en.wikipedia.org/wiki/Supernormal_stimulus

https://en.wikipedia.org/wiki/BLIT_(short_story)

Also: one of the V3A animations reminds me loosely of things I saw when I was a kid, at night, shortly before I slept (though my experience then was more circular).

If future generations of researchers will wonder why IRB reviews became mandatory for computer science, studies like this will be the answer.

Seriously, some people don't seem to realize the point at which they are becoming Fritz Haber.

e.g. cartoons from the 1950s were designed to have:

- cute animals

- short and fast scene cuts

- catchy music

b/c all three of the above were known to trick the human brain into "sit still and observe" mode.

Focus groups did the same thing for advertising which then morphed into A/B testing to determine revealed preferences.

Same thing with the whole fat/salt/sugar research that led to near addictive fast food.

This is just the next logical step.

But for the paper itself, it seems they're using genetic optimization over predefined keywords. Wonder what would happen if they did gradient descent on the latent space directly. Is brain stimulation just not a good domain for GD?

Literally hijacking people’s neural activation pathways on a mass scale that can’t be easily blocked, unless you look away or start meditating.

Are there any other precedents for this kind of neural-subterfuge being released into the wild, effectively lowering the barrier to mass manipulation?

I’m always skeptical of these papers that claim to read minds, because my understanding is current brain reading technology is very coarse (Neuralink is SOTA, very invasive, and only 1-3 thousand electrodes for 55-70 billion of neurons; fMRI is much worse) and neural networks generate plausible images and videos from noise. Indeed, the example videos don’t seem particularly stimulating, just random.

reason I am asking it could be some relief to our brains after tedious working day, especially after heavy AI usage

I can imagine this has some very undesirable usecases if used in the wrong way.

if it is targetting visual regions of brain and I have aphantasia (I cannot visualize anything in my mind) is that connected?

I don't think it's a matter of if but when. Grim.

[0] https://corticallabs.com/cloud

But here we can start also the usual discussion about technology research for the sake of it vs calibration of possible side effects of new research

Personally i think we haven’t solve this problem and thus it’s just a matter of time until we’ll get in a non-going-back point

https://www.afternow.io/content/images/2016/10/virtual_reali...

What could possibly go wrong?

Tolls like this have the potential to make this so much worse.

Will be interesting to see how strong the controlling forces can be - enough to make you miss things in direct perception like in the book, or only softer effects further up the cognition layer stack

I wish we had the Hippocratic oath for STEM, or at least that they would take ethics seriously rather than an afterthought against the god of Progress at all costs.

Is my brain different or am I just a grumpy millenial hipster?