I couldn't disagree more, humans will always be a liability and as machines get better they should eventually have total control.
Machines are only as reliable as the situations they were designed for, and their ability to recognise failures.
There have been a whole host of crashes/incidents that were caused by things that hadn't been seen before, and that humans either managed to work through and avoid them becoming lethal, or managed to minimise the loss of life by doing things that were not in any manual and sometimes went against all normal practices either due to the pilots own prior experience in other methods of transport, knowledge of the equipment, or sheer calculated spur of the moment desperation to try something, anything that might work.
Don't forget that there have also been a bunch of accidents caused by the computer systems in the aircraft deciding that two of 3 sensors were right, when in reality they were faulty, again IIRC several were averted by pilots doing things like using mk1 eyes and ears to diagnose problems
There are pros and cons to pilots vs computers for flying aircraft, but one of the massive pros of an actual human pilot is that he/she can cope better in a lot of emergency situations than a computer. The Hudson river crash for example is a great example of that, the pilot recognised what had happened, recognised that he had no chance of doing the "normal" thing in diverting and immediately started to prep for the only landing that was available which would minimise external causalities.
I don't think it's been talked about much, but Sully's decision to ditch in the river not only ultimately avoided the very high chance of massive loss of life on the aircraft, but it also ruled out the almost certain significant loss of life of people outside of the aircraft (the area was built up enough and he was low enough that almost anywhere other than the water would have involved hitting buildings).
I would also point out, that apparently a lot of the "aircraft behaves like x" behaviours that the computers rely on are, especially on older aircraft based (initially at least) on relatively limited real world data, especially for non standard situations. The aircraft manufacturers are oddly adverse to doing extremely risky manoeuvrers on manned airframes that cost several hundred million dollars each, so in an emergency situation a human pilot can be the first person to try (out of desperation) certain things that the onboard computers would be programmed to never do, sometimes those desperate measures work, sometimes they don't, but at least the human pilot will try something.
The current situation where you have a human pilot able to overrule the computer (at least when the pilot knows what the computer is doing *looks at post MD Boeing*), and the computer giving warnings is about the best option, unless the computers get much better and much more versatile whilst still maintaining at least the same level of safety.
No matter what you do there is going to be a problem with human error (or malicious acts), either in the cockpit, in the maintenance of the aircraft, or in the programming of the computer, and at least with two humans in the cockpit any mistake or deliberate act is going to have to get past two people + the computers.
