Dare I point out that humans also create the software?

Heh, Bitt, I expected this

People are not very good at split-second decisions, keeping focused on boring task for long time, judgment free of preconceptions, pressures etc; our sense of orientation, acceleration and turn rate in 3D is easily fooled; when judging speed and distance we automaticaly use cues that might or might not be appropriate... In all this, essential for safe operation of aircraft, machines can be made to perform better (since the problem space is relatively small, predictable and easily defined).

Of course, being written by humans, aircraft control and/or navigation software will contain errors. My point was that, given enough time and resources, software (possibly in complex setup like multiple-redundant 'voting' systems with different, independently designed and implemented hardware and software) can be made arbitrarily close to error-free. I think we are near (or perhaps even past) the point where cost of such a system (plus the cost of residual errors) is lower than the cost we are willing to pay for human errors. This cost of human-replacing control systems is, of course, non-trivial: couple of hundred thousand bucks will not cut it...