This post is a bit all-over-the-place, but I'll bite. I used to share this belief once. The only gains to be had from intimately understanding underlying structures are to be found in performance optimizations. You don't model a problem space by thinking in terms of what values go into what registers. If that were the case, we'd all be writing web applications in Assembly. Abstractions help us to reason out a solution without getting lost in implementation details. As Dijkstra puts it [0], "In this connection it might be worth-while to point out that the purpose of abstracting is not to be vague, but to create a new semantic level in which one can be absolutely precise." This is why the whole idea of Domain Specific Languages is so interesting. Any loss in flexibility is more than compensated for by the gains made in expressiveness when you can treat low level entities as black boxes.

[0] "The Humble Programmer" - http://www.cs.utexas.edu/~EWD/transcriptions/EWD03xx/EWD340....