Conversations with an AI - Theory of Intelligence, Self & Compression
Revisiting some hashed out things already. But. I had a bit of a revelation the other day. That neural nets, our brain, LLMs are just a way to store compressed data. aka I'm not special. No one is God's special child. The idea is this - If you have a system that has a "need" to store information about itself and its environment, how would it do that. At a naive first pass you can store on a 1:1 basis. Each piece of information you get. You store. But. This is almost instantly unserviceable. It doesn't scale. A brain store to hold a set of information would be so large to be unmoveable, not to mention its calorific cost would exceed any lifeforms capability to obtain that. You cannot store information at 1:1 If you are going to store information it must be compressed. One way to do this. Push information into a higher dimensional space. Imagine. 100 bits of information in a line. To store 100 bits of information you need 100 bits. 1:1. Let us push that into ...