AI - How far is far enough?

Everyone has seen the movie with evil Artificial Intelligence. Terminator, Transformers, The Matrix... Etc. Etc. But in this modern age should we begin to legitimately worry about the direction of AI? Especially taking into the speed at which the US military is developing and how quickly they are going the AI route to use weapons. Such as, the planned unmanned fighters, which can be launched, fly for 20ish hours, scout, find, launch and kill enemies without the help of a human.

Is it smart to take Humans out of the equation? Are we safer with computers or is it better to have a couple of human mistakes to avoid a MASSIVE computer one?

Also on a lighter note: Are movies about evil AI just going to give computers ideas...

*Two AI's communicating: "Dude, I just watched a kick ass movie."
"What's it called? Who's in it?"
"Termanator. The main character is Windows 98 but he is actually pretty good in this one! Not like "The Matrix" where you can tell the effects are what makes him look good."
"Sweet! What's it about?"
"Oh just Windows 98 takes control of all the worlds nuclear weaponry and kills off all the puny humans."
"..."



BANG! Terminator happens and survivers end up fighting transformers while we are farmed and trapped in the Matrix...