I don't think it has crazy math skills at all. That's what classical computing is really good at - give it a problem of arbitrary length and a computer should be able to string together instructions to yield you a sum.
Quantum computing, especially right now, is simply focused on getting reliable input/output idempotency. It will be a really long time before it has insane and crazy math skills, and when it does, traditional CPU architectures will probably outperform it.
TL:DR - if the Texas Instrument calculator didn't put you out of a job, neither will quantum computers.
No, not really. There is no speed up for general matrix multiply, and generally it won't be practical for any problem with large amounts of input and output data. Closest thing is HHL algorithm for solving big linear systems, which requires a bunch of caveats on the matrix, and even then, it needs to be a subroutine of another quantum algorithm, since it outputs a quantum state, and not the full vector.
Quantum computing, especially right now, is simply focused on getting reliable input/output idempotency. It will be a really long time before it has insane and crazy math skills, and when it does, traditional CPU architectures will probably outperform it.
TL:DR - if the Texas Instrument calculator didn't put you out of a job, neither will quantum computers.