michael_nielsen

A striking instance of this is de Groot and Simon’s work, where they estimated that expert chess players typically have internalized 25,000-100,000 different types of piece configuration. This lets them understand the board in ways completely inaccessible to beginners.

OK good point. Still, I don't think that's the generic case. And even with chess, you get to recognise configurations if you love analysing games.

Incidentally, & anecdotally, until pretty recently I thought of understanding and memorizing as rather more separate. After I started experimenting with memory systems, I was shocked by how much faster I started to understand things.

My eventual conclusion was that I had earlier greatly underestimated the role of memory in understanding. (This is much less true in areas I’m already an expert in; the effect is strongest for areas a little apart from things I know well.)

Can you recommend one that would convince me? I'm sceptical but I could try it on a thing I've been trying to understand that's a little apart from things I'm expert in.

Probably not. Learning the memory system was, itself, not at all trivial. Though it’s now repaid the effort 10-fold. But some basic thoughts here:https://twitter.com/michael_nielsen/status/957763229454774272 …

It was the experience with AlphaGo / deep reinforcement learning (described in that thread) that started me wondering if I’d been underestimating the role of memory. I’ve since confirmed that many more times.

Somewhat to my chagrin. When teaching people things like (eg) elementary quantum mechanics, I often feel that they’re handicapped by not really remembering basic terms, notation etc. They think they’re struggling with high-level understanding, when really they just…