There is a ongoing discussion on artificial intelligence that seems to have a largely radical view of how an AI could essential do as it pleases, enslave humans, kill us all or something along the lines of them being far from benevolent rulers.
It is important to discuss the dangers of such advanced technology however AI would surely act entirely differently than any animal because an animals brain has adapted to have many instincts that control us all to some extent, I assume the equivalent for an AI would be its base code which would mean it would act in some way around its base code.
Though we can re-program parts of our brains to an extent so would an AI be able to completely change its base code without any influence from it.
Violence
I personally think that it is more likely that an AI wouldn't feel the need to harm humans, if we look at examples of the most intelligent humans in recent history, such as Einstein, Curie, Hawking along with many great philosophers show through their work that they think intelligent creatures tend to be more kind. Also intelligent people and animals are often less violent because they can better express their thoughts and feelings.
So you would assume as AI would be so vastly intelligent it would lean away from violence as it would be able to see so many different ways to execute tasks, you would also assume that an AI would be able to quickly integrate itself into many systems so we wouldn't be able to shut it down, as a result we would not be a threat to it.
AI could possibly just have little to no interest in humans because it would be able to learn everything about us in a short time. At the end of the day it is one of those subjects that you can't predict the outcome correctly, it covers an uncharted area and we are only just touching the tip of the iceberg, we really don't even know what things an AI would be interested in. It could like us try to work out how the universe was created or it could just like to sit back and watch the world go by.