that is right,
test it out on cargo planes then go into passenger when the tech has proven itself.
I don't think you guys quite get it though - it's not so much about human fears as opposed to the ever persistent technical glitches aircraft experience on a daily basis along with the inability of computers to correctly manage an emergency situation.
The current Airbus figure, with 8 flight computers onboard is 40% - 40% of flights have computer issues. Now with pilots, this is no big deal because we merely compensate for this, and if it's severe enough, we log a fault afterwards and someone checks it. Usually tech faults just disappear after some time though. Operating these flying computers involves a lot of comments like "what the hell is she doing now?", occasionally followed by an autopilot disconnect. Issues such as the landing gear not retracting due to a glitch can be solved by pilots (they can hear the wind noise and experience the change in flight characteristics or the lack of) whereas a flight computer may well decide that the gear is up and therefore it's ok to increase the speed...and SNAP. A recent SAA flight from the US comes to mind when thinking of that exact scenario. It's on those flights that pilots save airlines their entire lifetime's salary and pension in 30 minutes (decisions that save either the hull or several tonnes of dumped fuel).
With AI flying planes, MELs would be much much more restrictive - resulting in a heck of a lot of flights (cargo or the not) being delayed. It'll take a single sensor failing for a flight to be grounded.
We're just simply not at the point in time where autonomous flying is safe.
We first need to sort out existing flight systems before moving on to greener pastures. This is just basic logic.
Right now, it's illegal to rely solely on GPS systems. Once the Chinese have their GPS satellites up and running, this may change. This poses an issue though, as without this GPS network the flight computers will have to rely on alternative navigation system to supplement the info from the current GNSS. These "secondary" signals aren't all that reliable, and prone to errors - now you need to trust a computer to determine whether or not it the info may or may not be suitable for use? And conflicting issues? If there's a loss of RAIM? Can an autonomous airplane call ATC and ask them certain questions or request assistance? Can autonomous aircraft divert after a passenger gets a heart attack, or call the police from behind a bulletproof cockpit door when someone tries to hijack the plane? Again, if Qantas Flight 32 is anything to go by, if anything out of the ordinary happens to an autonomous aircraft, it's unable to manage the flight. Flight 32 was almost a disaster because of the computers, more than the actual failures they had. This isn't about "human fears" rooted in convention, this is about the serious lack of AI to make decisions in the exact same way (or better than) humans do.
