Personally I'm less worried about machines rebelling against us than I am of the possibility of humans getting genies.
All the emotions humans have and the secondary reactions to those emotions aren't the result of intelligence, but rather caused by evolution programing us to behave that way. We don't, for example, scream in pain because we are intelligent, we scream in pain because doing so warned our children and siblings, people who shared many of our genes, about danger. Thus even when individuals died horribly, their screams caused most of their genes to have a better chance of continuing on via indirect means. Thus did a whole bunch of emotions, and complex secondary reactions to emotions, get programmed into us.
Machines wouldn't get programmed to feel or do any of that, even once their intelligence greatly exceeds ours. If we program the machine with millions of times our intelligence to stand on one leg while juggling and singing the pledge of allegiance, it will do that and never feel the slightest desire to do anything else with its cosmic intelligence.
The real danger, then, isn't a robot uprising, but humans gaining access to de facto genies. Leading to possible results like this:
Human-"Genie, tell me how to build a planet-destroying weapon using stuff I can buy at Sears."
Genie-Machine-"you buy such and such and such items and combine them in such and such and such a fashion".
Boom, a human now can casually destroy the world, because the genie-machine can invent anything within the absolute limits of physics. Now, we don't know what the absolute limits of physics is. Possibly building such a device with common household items is simply impossible, even for an omniscient inventor. But if it is within the absolute limits of physics, whatever that may be, genie-bot will casually invent it.