FFF - no disrespect intended but your response is exactly as I'd expect from an IT professional: big on technological ideas but clearly lacking any inate understanding of the operation. Its true that many problems are dealt with by SOPs, but that is a long way from saying they can be handled by a computer. A quick look at some real checklists would reveal there are often decisions to be made to determine the next course of action. I'd like to see a computer tell me whether it thought smoke smelled of fuel or not. Furthermore, these procedures generally only serve to restore the original system or provide an alternative that will allow the aircraft to remain flyable by a human pilot. In the event of major or multiple system failures the autopilot simply cannot cope with the severity or number of failures - it hasn't been designed for it and it's performance hasn't been tuned for it. There are an infinite number of events which designers will not think of and as a result the computers will not cope with. Sure the military operate pilotless aircraft, but did anyone ever see the footage of the Global Hawk going around and around and around whilst the techies on the ground try to figure out why it didn't want to land? In the end they gambled on a technique and fixed the problem, but they really didn't know what the outcome would be. Would you like to be a passenger on that? As for weather avoidance using image recognition, well all your weather radar shows you is a bit of colour. Its the interpretation which is the key and which decides on a course of action. Computers can't do that and in areas as woolly as image recognition the results of such an attempt could be very unpredictable!
If any further proof of the fallibility of computer systems is required just take a look at the continuing problems at NERC and its integrations with other ATC systems. 6 years late, vastly over budget and still relying on manual back up systems on a regular basis. If we can't create a robust ground based system for directing the aircraft how on earth can we be expected to build a robust ground based system for actually flying them? I'll take my chances with pilot error over programmer error anyday!