I'd like this to be true, as I want the time I spend learning math in the future to be as useful as you say, but I seem to have come rather far by knowing the superficial version of a lot of things. Knowing the actual math from something like PT:LOS would be great, and I plan on reaching at least that level in the Bayesian conspiracy, but I can currently talk about things like quantum physics and UDT and speed priors and turn this into changes in expected anticipation. I don't know what Kolmogorov complexity is, really, in a strictly formal from-the-axioms sense, nor Solomonoff induction, but I reference it or things related to it about 10 times a day in conversations at SIAI house, and people who know a lot more than I do mostly don't laugh at my postulations. Perhaps you mean a deeper level of understanding? I'd like to achieve that, but my current level seems to be doing me well. Perhaps I'm an outlier. (I flunked out of high school calculus and 'Algebra 2' and haven't learned any math since. I know the Wikipedia/Scholarpedia versions of a whole bunch of things, including information theory, computer science, algorithmic probability, set theory, etc., but I gloss over the fancy Greek letters and weird symbols and pretend I know the terms anyway.)