Just a quick little page to get some ideas out of my head.
- Because we have a very low-bandwidth communication medium (speech), communication between people requires each person to have a lot of contextual information in their heads so that they can transfer data quicker (i.e. you must both speak a language, you both need a similar set of common experiences to be able to relate about things or draw analogies, you need to remember a large chunk of the entire conversation up to this point, etc.)
- You could say that, to be able to communicate more efficiently, we have to compress the data in our minds first. Since speech is so low bandwidth compared to what's going on in our heads, we have to use a very lossy compression algorithm to avoid each idea taking an hour to say. Also, compressing the ideas is important because we don't have a very large short term memory. If you're explaining a complex concept, you have to keep the concept in mind as well as all the words you've currently said, so that you know what the other person has heard so far. And, on top of that, the other person then has to keep all those words in their mind as well, WITHOUT the aid of the clear and detailed concept that's in the first person's mind. Lossy compression, or approximation is necessary due to our own limitaions.
- An example of this data compression is when you know somebody so well that nobody else can really understand when the two of you converse, since you speak in a kind of verbal shorthand.
- Encryption is also like compression, except with encryption, the context that you need to decompress the data is a single value -- the encryption key. A highly compressed conversation could be considered to be encrypted since there is some context or "key" that, if you knew what it was, would make the entire conversation make sense.
- Learning is also a kind of data compression. When we learn something, we're basically finding patterns in things. To learn what a chair looks like, we see lots of chairs and find the pattern in it. (eg. in the abstract, a hair has: 4-legs, a flat surface slightly larger than a human ass that's about as high up as your knees, it has some kind of back, etc.)
- Finding patterns allows us to compress the information that's stored in our neural network.
- When there's a simple pattern that underlies many things, we save a lot of neurons because there only has to be one representation of that pattern which everything else is linked to.
- Scientists are quite certain now that REM sleep is used to compress and filter the information that's stored in our neural nets. REM sleep has two purposes: 1) it wipes clean all of the random and insignificant patterns that we learn during the day, and 2) it retains patterns that persist from day to day. This is why we forget things we don't access the memory frequently (by thinking about it, or re-experiencing it, or even just thinking about something that's tangentially related to it and which has associations into it that memory which activate the neurons slightly (that's what an associative memory is)).
Strangely, dolphins have proved that REM sleep is data compression -- dolphins have larger brains than we do, yet they're nowhere near as intelligent. It was puzzling at first, until scientists discovered that dolphins don't do REM sleep. Since they have no way of compressing the data in their brains, evolution came up with a very elegant brute-force solution -- make their brains bigger!
- Our minds, despite being hugely complex, are actually quite feeble. It's very lucky that the universe has a lot of simple patterns in it -- it makes it quite easy to understand. That makes us feel smart and powerful, even though we're just tiny little specks of dust in the cosmos.
Other interesting things: