Ask a rock what's 2+2, it won't be able to give you any answer. But ask a computer, and it'll tell you 4.
No. Unless you did install a voice recognition software and a microphone and…, but that's not the point. The point was that the rock could be seen as some kind of
analog computer, only one that does only compute something we are not interested about.
Now let's take the example of another kind of analog computer : a abacus. Is the abacus intelligent ? Of course not. It's an inert piece of wood. Can it help you do computations at least twice as fast ? Yes. It's a good tool. But it's not intelligent, it is merely designed to help you in your intelligent process. It's the same with the computer, only the tool is more complex. More complex, but just as inert.
A rock has absolutely no method of communication
Not true. It can communicate by, say, it's speed. For instance, if you do wonder how fast one particular rock would fall if it fell from, say, Big Ben. Then let me go up to Big Ben, let this particular rock fall on you, and it will indeed compute (in an analog computer way, which is in no way less valid than the numerical computer way), with an infinite precision the speed at which it must fall, and then give you the answer, again in an analog way. It's true it won't give you the answer numerically, but it will give it analogically. In no way is this less valid. And if you really want some numerical answer, but something at the bottom of Big Ben to measure it's speed and convert it to a numerical value.
It wouldn't be a computer otherwise, as you wouldn't be able to give it commands.
Why, some “computers” are build specifically to solve one, and only one problem. They are not all general purpose computer. And generally speaking, a computer that would be designed to solve only one problem, with only one fixed input, could be made much much much more efficient than a general purpose computer. I can try to find examples of such kind of computers being actually build.