Human Memory

I've read an awesome book recently - The Design of Everyday Things by Donald Norman. I must say that didn't expect much of this book. My thought was "just another book for designers on how to create usable things". There was however something pushing me to read this book, maybe because  I've read about it in the In the Plex; book was references as the one that influenced Google's founders. Well, now I can understand why. Even thought the book is mostly about trivial things that everyone should understand and know. In fact, it's not true. Not everyone understand and know. I didn't. So many openings about regular things, views from different perspectives, inspirational rules etc.

There are many topics that I liked in this book. But the chapter named "To Err is Human" maybe the most favorite for me. Not only because I make so many mistakes and errors all over the time by myself, and it's nice to understand how this works (and how I work). But because author gives very good explanation on how human memory and brains work.

I made some notes during reading this book, and decided to share some of the them that are related to how human memory works. I also was thinking how this apply to the AI. And made some interesting openings for myself too.

So here are my notes...


Memory could be seen as a graph of different facts and experience. Memory could also be based on intelligence and deduction, ie if A and B then C. Thoughts and memory are closely related.

By thinking we can reorganize and improve our memory structure, create a new relationships in the memory graph, that is different from what others have, and be more intelligent and knowledgable. We could see more details and aspects, thus be better at finding solutions, hidden relationships etc.

That's why repeating information or gaining it from different sources, increases our memory and understanding of many things. We basically create more connections in the graph, thus improve our memory and can think even better and faster, see more relationships etc.

Other than that, human memory is very chaotic and messy, very often we have strange connections between facts. Subconscious could play some role in organizing our memory, by filtering non-important facts and ignoring them or creating weak relationships to them.

Our brain tends to generalize experience, but at the same moment exceptions have larger priority.


It is very important to make mistakes and learn on them. For example. for kids or AI there should be some priority to detect mistakes, solve them and learn to become experienced.


Our subconscious can keep focus an many things and process them in parallel, while conscious only can keep focus on single thing.

Thus subconscious seems to be way advanced than conscious, but it has limit of control. It can however send weak signals, or be well heard only in critical situation or when conscious is looking for advice. Subconscious could play a role of the intuition, as it handles large amount of data, use known and messy techniques trying to handle the incoming huge set of information.

Parallel distributed system

Our brain is a parallel distributed system, with many independent or weak-dependent systems working in parallel. Some systems work on background, like our subconscious, others work in foreground like our conscious, some work in foreground but in parallel, like our motion/vision/hearing mechanism.

If we talk about AI, then I can imagine that it will consist of many many different distributed systems, each taking a huge amount of data and should be able to handle them. Subconscious and conscious are both required, and consist of many smaller distributed systems, each executing single or multiple simple functions and impacting AI behavior.

No comments: