The typical view of
nanobots given in
films is that they will be able to do practically anything with little or no
input in the way of
fuel or precice control.
The problem with a nanite is that if it were to act totaly independantly it would need a very sophisticated internal compter to guide it and control it, which is obviously very difficult in a machine less than the size of one of our cells. A more likely aproach will be a client/server architechture, in which a massively powerful processor will balk commands via microwaves to each individual nanite or group of nanites, giving them very simple comands based on their environment and interactions. The main problem here of course is that a processor powerful enough to compute all the necacary commands would need to reach into the terahertz range to controll any more than a few dozen nanites.
Intel have predicted that in ten years the gate of a transistor will consist of a single molecule of silicon dioxide, or three atoms. This is the peak of all possibilities for silicon dioxde, and if computing power is to exceed this point, a new medium must be found to create transistor gates. The most advanced processors Intel predict will have been produced by that time aproaching the ten gigahertz range and running on a mere three volts. If this is the farthest we can reach in the age of the superstring and the quanta, how long will we have to wait untill we have gained sufficient insight into the physical world to create even faster computers capable of the unthinkably complicated tasks we place upon them?