Sorry, if this has been mentioned before, I haven't gone through alle the replies.
One thing automation won't resolve is human error. How many of those "human errors" that lead to plane crashes have actually been errors in using automated systems?
I'm thinking of that case where pilots accidently entered a wrong waypoint into the FMS and the plane flew into a mountain. Or the case where pilots accidently entered a 3500 FPM descent instead of a 3.5 degree descent and the plane flew into the ground. Or the case where pilots accidently entered a heading of 020 instead 200 and ran out of fuel over the jungle.
In all these cases, the planes were flying on autopilot and the human error was not in handling the plane, but in entering wrong data into the automated systems. That will happen in fully automated planes as well.
There's one thing that humans can do but automation can't: Humans can recognise their own errors. A human being can make an error, and a second, or a minute later recognise their own mistake and correct it. An automated system can't.
In the three cases I mentioned above, the pilots didn't recognise their own mistakes in time to prevent the accident, but we don't know how many similar cases happened, in which pilots accidently entered something wrong into their FMS only to spot the error in time to prevent anything bad to happen. And I truly believe that a pilot who is actually there and will be the first to die in an accident is more likely to be vigiliant and spot their own erros than a ground operator who is far away and is possibly distracted by operating several planes at a time.