Leave closed-loop artificial intelligence for solving engineering problems like smart bombs, railway yard traffic, nuclear power stations and kitchen refrigerators with 17" plasma screens bleating interactive instructions on how to use a broom, all of which have produced catastrophic outcomes that are totally blameless.
Your golden tongued rhetoric does little to advance the discussion. First of all, servo loops do not in any way fall into the commonly accepted category of artificial intelligence. They are fundamentally furnace thermostats with various forms of dampening and anticipation. In the Quantas incident, the servo loop was not responsible for the injuries, a defective sensor was. Granted, flaws in the system prevented the crew from identifying and isolating the defective subsystem, but that will be fixed. Secondly, any pronouncement of unsuitability should also include a risk analysis of other ways. My risk analysis tells me that in the USA, the major airlines have flown for nearly 7 years (knock on wood) without killing a single passenger. That tells me that something is working right and I'd rather not go back to whatever way it was before.