In most instances, the various paths cancel out so it looks like it went directly from point A to point B. But in the two-slit experiment, each particle takes paths through both slits, and the way the paths cancel leaves the two-slit pattern. If you instead can detect which slit the particle has gone through, then you've eliminated the paths through the second slit, and you end up with a single-slit diffraction pattern.
It's not just TWO places at once, it's INFINTELY MANY places at once. And the way that we "prove" this is that the theory matches the behavior that we observe. I realize that this seems counterintuitive, but if you hope to understand quantum physics, you will often have to set common sense aside.
For a layperson's treatment of this topic, get a hold of the book, "QED: The Strange Theory of Light and Matter", by Richard Feynman.