I appreciate your argument, but I feel like you missed an important motivation behind my question. You see, I'm worried about unknown unknowns. Once you know enough about a subject to choose an itch to scratch, you don't have to think about the difficult foundational subjects like compilers and AI. But before you get to that point, there are too many unknown unknowns. Like Yegge's perennially disappointing interview candidates, we can't say "grep" instead of "a 2,500-line state-machine in C++" if we don't know that grep exists. It's an unknown unknown, and while I know what grep is, I don't really know what happens under the hood when it comes to syntax trees and machine learning, and since they are foundational concepts of my field, I feel like I need to dive into them as soon as I can, to make everything else I do more informed and efficient.

In that case, read, but don't try to understand in full and complete depth, a few books. I would suggest:

* SICP

* The Dragon Book

* Programming Pearls

* The Elements of Computing Systems

There's a shed-load of good stuff, and no doubt others will suggest more.