Obviously, the problem is in the definition of "to think".
Some people think that to think, is to have some internal word-based discourse.
And indeed, when you are learning, eg. mathematics, you are transmitted a formal discourse system, such as two plus two equal four, eight plus five equal three carry one, plus some rules such as: add digits and carries from the right, to the left.
But notice already the limits of this system: even when you're a young adder, you won't keep telling you the rules: "add digits and carries form the right to the left", or current adding column is the second from the right, let's move to the next column on the left.": you JUST DO IT.
Ie. once you've learned the rule, you apply it without thinking. (I mean, without this kind of verbal discourse). Once you've learned the tabular data, you also use it without needing to recite it verbally.
This is the same when you learn music, cooking, bicycling, or anything. At first, you have some verbal description of what you should do, step by step, and you may "think" about it to interpret the "program". But if you really learn it, your brain will "compile" this verbal program into your subconscious.
And this is why it is important for children to learn mathematics reasoning and programming, not because they'll become mathematicians or programmers, but because they'll learn reasoning processes that their brains will compile and that they may forget the sources (the verbal lessions given to them). But their brains will have been hard wired with those compiled logical reasoning programs that will make them and the world better.
Oh, and yes, when you know programming, your verbal centers are free, so you may use them for other things, such as temporary storage "line 38, line 38" :-)
This is also a reason why I'm not sure vocal interfaces could be made quite usable for programming (or for mathematics, or polytonal music, or for sports (try to tell a robot how to move each of its muscles to jump over a bar)).