Right, as he says later: [...] In context, the quote is more about verbal overshadowing and related biases, wherein having a map can blind one to the territory, and the excesses of academic tail-chasing and status-bound disdain for the merely practical. In other words, it's rather a lot of things lumped together, each one of which has been an OB or LW post topic at one time or another. (Which is why I thought it appropriate to link to the whole thing, rather than just giving one out-of-context quote.) [...] Wasn't the effort involved in generating the practical knowledge of how to build a nuclear weapon at least a couple of orders of magnitude greater than the effort involved in coming up with the idea, even if you count all of the physicists in a direct line from Newton to Szilard? Part of Taleb's point is that even if you have a promising theory, you are really only just getting started -- and then only if you don't have a model that blinds you to the real thing. And for a great many things (especially those where fast feedback is possible), you will get better results sooner by building your map from the territory than trying to come up with a theoretical model from scratch. One reason why knowledge doesn't flow equally in both directions is that theory is a more compact, "lossier" form of information that is necessarily included in practice. Another reason is that human brains are better at building intuitive models from experience than from principles. (i.e., better at extracting principles from experience than generating experience from principles.)

Exactly. And it works the other way round also. If you know that throwing seeds into the ground will yield plants you can gain theoretical knowledge from that. Edit: I think my example was not that good. I'll give another one: if you can fold paper planes that fly you can gain theoretical knowledge for building a real plane. This is what models are for.