logoalt Hacker News

MarkusWandelyesterday at 10:42 PM2 repliesview on HN

I think that's because memories really are highly compressed, with the average complex memory (of a grand day out, for example) being compressed into at most a paragraph of text's worth of information. Everything else is made up of context.

For example, having no photographic memory, I can remember the general "feel" of the innards of an antique shortwave receiver I used to have as a kid. But almost all of that is context. I know what canned octal base vacuum tubes and big tuning capacitors look like, what sort of components were in the big band selector "sled" (for geeks: This was an NC2-40D) and so on. So I can "zoom in" on the coarse memory by just filling in details that are generic, not specific. And these details may be from different sources than the original memory too.

And so the brain needs to remember - if one is not faceblind, which in fact I am - just the overall geometry of someone's face, and all the other stuff can be filled in as needed (how they look when they smile or are anxious etc.) Compression, compression. We have a lot to learn from nature. Note that a faceblind person does this too, of course - just not with a geometrically exact visualization of that person's face, just one with the correct "feel".

And that's also why synthetic memories can be formed so easily. You really don't have to insert much. The brain infers the rest as needed.


Replies

dwdtoday at 2:05 AM

Given the number of neurons, hash collisions are inevitable? /s

Our model of the world, and this includes "memories" is very highly compressed, and extremely lossy. We reconstruct everything from what neurons activate when we recall something, so it can also be like a corrupted zip, where we don't access everything required in the moment. This can be affected by sleep, stress, hormones, etc.

What is quite interesting is when we try and recall something and can't, but the subconscious processing continues and you recall it later.

MarkusWandelyesterday at 10:47 PM

This of course applies to my (sarcastic) description of women's memory too. A given conversation is compressed to a small amount of memory, a significant part of which is the "feel" of the conversation rather than the actual words. So of course the memory is not word-for-word exact. But, let's just say, a subset of persons, a majority of which may or may not be female, does not acknowledge this and insists that the (reconstructed) sequence of words is exactly what was said 20 years ago.

show 1 reply