I agree that it seems unlikely that we'll achieve nanobots that are more complicated than biological cells. An example of an unrealistic idea would be a remote controlled (or even autonomous) submarine diving through blood vessels, broadcasting what it sees to the outside world in HD quality.
Still, lot's of interesting things should be possible...
Such devices are already possible because the scale needed is mm not nano. For example the devices to do angioplasties are pretty big because blood vessels are pretty big. There is also a lot of commercial and research work being done on "remote control" treatment delivery, esp. for cancer treatment. (By remote control I mean externally guided as opposed to joystick control.)
Still, lot's of interesting things should be possible...