Hacker News Reader

Comment by phkahler | Hacker News Reader

This fits with (not sure the term) the model of our brains constantly doing projection/simulation. Your consciousness is at the abstract level and the brain projects successive levels of detail, then compares to actual input at the lower levels looking for differences to correct the simulation.

In the abstract this is very similar to how a Kalman filter works. There is a prediction component and a correction, with a (possibly time varying) adjustments for how much weight to give to measurements. The equivalent here would be to simply ignore the measurements when the eyes move since the input will be known to be messy, and just rely on the prediction.

This turns out the be a very effective (and in some ways mathematically optimal) way to process sensor input even in human-engineered systems.

View on HN

Right. We are constantly sampling the environment in small amounts, but we perceive a much larger, more stable and detailed space than we can directly sense in recent moments because our consciousness is based on a simulation integrating a lot of information already gathered.

https://youtu.be/NTnbnw01HHY

I think the author of these tweets is falling in to a common trap of trying to reverse engineer why our brains do what they do and try to ascribe intent to randomness. Everything our brains do is the product of a billion years of random evolutionary events leading up to a system that does something useful for our survival. Trying to explain how our brain processes "images" like a camera is the first mistake. The universe is a wild cacophony of matter and energy and what we see in our brains is just a rendition of the salient bits of data gathered from that noise. Looking at a second hand on a clock or a mirror was never an evolutionary selector. Neither was seeing infrared or ultraviolet. At least, not for humans.

If you want to compare it to a system, then a neural network is actually pretty apt (no coincidence there). Data goes in, conclusions come out. The mechanism takes thousands of byzantine shortcuts to get from A to B and the result is a small number of egregious blindspots in an otherwise highly efficient system.

This fits with my personal model of the brain composed of a distributed system made of millions of dumb organisms screaming in coordination to create a bigger, more intelligent being.

Kinda like an ant colony, CUDA and Kubernetes mixed together

> Kinda like an ant colony, CUDA and Kubernetes mixed together

Thanks for the nightmare fuel.

No shit. Excuse me briefly while I go calm down on the demonic spaceship from Event Horizon.

Reply on news.ycombinator.com