1) Computers don't have anywhere near the same processing power as the brain. They are many orders of magnitude better than us at one very specific task (moving numbers around). We have no idea to what degree "really fast number shuffling" factors into intelligence.
2) Why would we hook it up to the internet? We could easily provide access to whatever knowledge we felt it needed, but still leave it air gapped.
3) Why would it have an motivations whatsoever regarding humans unless these were programmed in? Our motivations are the result of billions of years of natural selection (things that want to survive are more successful than things that don't). A "created" AI would be motivated by whatever we had built into it.
4) Why would it necessarily be able to replicate? There is no reason to think it would understand any more about it's own intelligence than we understand about ours?
5) Why would be give it the access required to replicate? Unless we gave it permission and access to a text editor / compiler and execute permissions on the code it generated, and access to hardware to run it on, replication would be a problem.
Hollywood stories are all good fun, but stories about AI typically don't stand up to any sort of reasonable analysis. IMO, our first AI's will be like very intelligent friendly dogs, willing obey master in order to get the reward that satisfies whatever "need" we have built into them.
Treating it "poorly" is an interesting question. Is it unethical to "enslave" an AI if it has been designed to get the most satisfaction out of "enslavement"?