The pilotless aircraft has always been the goal of every aircraft. Their problem is manufacturers can't work out how to do it. Their first problem is existing software. It is so full of bugs it will give a plate of rotten seafood a run for its money. Many bugs are minor, but they still exist. When bug free software is delivered to existing aircraft then the software industry is ready for the next step. If Honeywell are involved, pilots will always be needed. We also have to change the nomenclature of intersections and existing navaids. A cleared direct to SPL, OL, LBA, COBOL, HAVEN might change the outcome of your day. Weather avoidance is another issue. Who will be learning which clouds to avoid, which tracks to take to avoid turbulence on finals? Academics? Experienced line pilots? I can answer that one easily - whoever is cheapest. Then we have the unexpected. Currently, carbon based life-forms do a reasonable job at self programming, failing that they regularly guess well. And a computer? Or will this be sub'ed out to a call centre in India?
Currently, aircraft system architecture is inappropriate for unmanned. They are too centralised and too cumbersome. We probably need to revert back to autonomous analogue systems for basic flying and digital systems for decision making. Then they have to be glued together and wrapped up in a totally bullet proof communications system with failsafe for when they fail.
And they are just come of the bits a soon to be replaced, bog standard, worthless line pilot can foresee.