Trying to reason about the brain like it's a computer is a very common thing for computer scientists to do. Unfortunately, it's also mostly fallacious (since for a whole host of reasons, neurons don't work like integrated circuits), and simplistic analogies like cerebellum = GPU are probably not going to give much insight into what's really going on. I'm not saying "the brain is beyond human comprehension" or anything silly like that, just that you have to approach it from a biochemical context.
Sure, but the brain is a Turing machine, albeit with hardware acceleration of key functions (e.g., edge detection in the visual cortex). Mathematics, not chemistry, structures the problems it solves, if not the algorithms and heuristics involved. It's utterly fascinating to see how nature tackles the same problems that we solve independently using completely different tools.
