Watson is, fundamentally, no different from that rock. Sure, it follows a very complex "path" indeed (though laid down by humans), but the only difference between the rock and Watson is the *kind* of path. In fact, Watson's path is less complex than the path of the rock (which isn't entirely a fair comparison, since the rock's path is practically infinitely complex)
It seems that you are assuming Watson is as deterministic (if not more) as the act of rock throwing. It is not. Watson is a lot more heuristic than algorithmic. We are only scratching the surface of what neural networks can do, and many times we do not understand why they behave the way they do. And we built them.