It seems to me that "pilot should always have 100% control" is becoming a fallacy. I think we should hand off emergency procedures to the computer completely. I think at this point, the systems are so complex that the computer might know better what to do than the pilot, given how many sensors it has access to.
Um, no. Do you know of [ame="http://en.wikipedia.org/wiki/Air_France_Flight_296"]Air France Flight 296[/ame]? While it might've had its share of "pilot error"-based causes, the computer was a significant cause of the accident as well.
The computer goes by just sensors; it has no idea of the situation. A pilot might have an idea of a situation, or he may not. The computer still is not much better off.
If your pitot tubes ice up at high altitude, the computer thinks "My airspeed is too low" and will tell the pilot. Most often, the pilot will mishandle the situation (ahem, AF447). The aircraft in question is perfectly fine; a AF447-type accident would only occur when the computer is like "This is too much for me" and gives control to the pilot, who then tries to remedy the situation, but then crashes the plane. The computer, a computer in general, would never be able to handle an accident or a anomalous event; that's why many autopilot can only pitch the plan up to 15 degrees (to name one parameter). Once they go beyond that, they'll disconnect. As in, give control back to the pilot.
Computers aren't designed to resolve things like accidents. That's the pilot's job.
The pilot might not do a good job resolving, but he's still supposed to. NOT THE COMPUTER.
Also, computers aren't exactly dual redundant. back to those pitot tubes, most planes run their airspeed data off just one. In more than one accident, the pitot tube got
up ([ame="http://en.wikipedia.org/wiki/Birgenair_Flight_301"]by a
mud dauber wasp once[/ame]) and the computer thought the plane was doing some thing it wasn't really doing, and then acted accordingly to remedy the situation (slowing the plane down because it thought it was overspeeding). The poor confused pilots tried to remedy the situation based on the faulty readings the plane was giving them. The plane stalled and well, crashed.
Frankly, data-checking between three or more pitot units, in where ones that are determined to be faulty are "voted out" (some s/c computers do this) might help prevent an accident of this kind, though it could just as easily cause one like this.
There isn't, and probably will never be, a perfect mix between man and machine that will allow for mistake to not happen. It's impossible, IMO. :2cents: