I would strongly oppose any view that says all accidents nowadays are due to human error, so we need to design the human element out of the cockpit.
These claims can only seriously come from people who have no real experience in real life commercial ops. As a pilot I make so many decisions on each work day to smoothen things, and deal with so many technical issues each month, I just don't see how we could ever construct such a complex machine like a modern airliner which will operate to a safety level of 1 in 10 million (that is where wer are right now industry wide) fully autonomous!
A plane is much more than only software and computing power. It consists of metal parts, mechanical engineering, interacting systems, and last but not least a whole bunch of sensors. Each of them can fail, which will pose it's own challenge to a solution.
And this whole amassment of technology operates in a real world environement, in challenging weather, with other participants who can whirl your own idea how to proceed in a second. If your company operates a good sms with publications, just read what kind of freakin stuff regularly happens no one ever would have thought about.
1 in 10 million fully autonomous? Never in my life time.
Which brings me back to the original question. We need good basic skills more profoundly distributed throughout our industry. TK in AMS, Asiana in SFO, AF in the Atlantic, Air Asia over Java Sea should never have happened. These accidents showed major deficiencies in piloting skills (not just stick and rudder, but also understanding what flying means in aerodynamic terms). I however believe strongly, we can train pilots so such crass things do not happen again. May be in the last 20 years, we traded too much protection for skill. But the industry is slowly changing so we add protection to skill.
There is still a lot to be further gained also on the technical side. ROW/ROPS systems hopefully will improve excursion statistics. We could possibly go further with braking coefficient measuring and making it available to following aircraft in the approach. GPS technology and therefore approaches with vertical guidance can further improve CFIT risk. And the whole complex of TO security has still a lot of potential, I think for example about TO data calculations not on your computer anymore, but directly within your plane FMS. No chance of a typo when entering speeds and TO power, plus the potential warning, when you are not taking of from the position you have calculated (GPS comes in handy again).
But all these will at the end need a pilot who can cope with system and sensor failures, and fly his plane by hand, with a good scan of the main flight parameters. Plus who can do all the tactical decisions that happen every day out there.