I am not sure I agree. The argument only seems to work if you consider a very simplistic teleological model of skill learning: that the only benefit in learning a skill is mastering a very specific task, e.g. multiplication, or derivation. But you learn a lot of other skills on the way that may be useful for completely different tasks - or even tasks which aren't known yet and thus cannot yet be trained specifically.
As my university teacher in Latin once said: sure, you don't need Latin anymore, but learning it will unlock areas in your brain previously unknown to you.