This is obviously all extremely speculative, but it's such an interesting topic I can't really keep myself from posting anyway. I've had this idea for the bett

This is obviously all extremely speculative, but it's such an interesting topic I can't really keep myself from posting anyway. I've had this idea for the better part of my life, that I try to avoid read especially large canonical works deeply, and with few exceptions prefer to read several sources shallowly to reading a few deeply. The exceptions are when I try to learn something entirely new, like a new mathematical topic, or when I encounter something that is both completely at odds with my current mental model, and not obviously false. From a learning perspective, it could be seen as to adding some kind of noise to input data, and to also have a kind of indirect - but rather strong - validation set through the knowledge contained in the canonical works, as understood by others. Could this prevent some kinds of over fitting? I honestly don't know, but maybe it could? It's obviously anecdotal, and extremely so, but it actually seems to be the case that my generalization abilities, my capacity to apply knowledge across disciplines, and to identify behavior on certain scale has had a rather unusual trajectory. Sometimes it almost appears as if I know things I don't remember ever having learned, or even though about. As if the "data points" have become so numerous that the brain somehow becomess able to generalize over things that one would (superficially) believe to be facts. Before I realized what was happening, I have had several people, some within highly skilled fields, come to believe I worked in their field after conversing with them for a while. When they realize I'm not, many became confused, and some to the point of being unable to continue the conversation. Nowadays I generally manage to avoid those situations, but it really was - and is - the strangest thing when it happens.