Human in the loop

As war begins to require faster processing speed and a faster reaction time, should we always require that a human gives the robot permission to fire, or should we begin to just define objectives and let robots fulfill them on their own?

To decrease cost and make the military more efficient, the military is planning on eventually have one or a small group of people controlling many, many unmanned vehicles. At this point, is it even possible for a human to effectively control the robots and process the information well enough to give the machine the ok to fire? Also, a human operator would probably be tempted trust the vehicle's judgement without waiting for his own head to confirm it, especially after the vehicle has been right several times. So if the robot makes a mistake, and a human still trusts it, whats the advantage to using humans if they still make the same mistakes, and can slow the robot down from acting on correct decisions. Because computers can process faster, it just might make more sense to connect the robots to each other, have them communicate the data and decide what to do as a group, but without humans slowing them down. Mistakes will happen, but they do anyway with things as they are. And if we can refine the robot programming, might they one day make even fewer mistakes? We may never want this to happen, but there are areas where the human brain might just become obsolete.