[dputtick]

Currently, pilots have the primary function of monitoring the automation, and exercising critical thinking when things go wrong. This can be everything from flight computers misbehaving, dealing with weather, avoiding turbulence, troubleshooting an engine fault indicator. Their secondary function of flying the plane is more or less because they have to be there anyway, and for almost all larger aircraft the autopilot takes care of everything except for maybe the first few minutes and the last ~1 to 10 minutes of the flight (although this varies based on airline, aircraft type, and individual pilot preference). Autoland does exist (google cat. 3 ILS), but it hasn't really been refined to perfection because the pilots aren't going anywhere as of yet and they are quite excellent at performing the last 200 ft of the landing sequence (once the runway is in sight in low vis conditions). Your average airline flight, in daylight and good weather, can really be quite a simple exercise. The outliers, with equipment out of service, bad weather, icing, or really serious things like an engine failure, are extremely cognitively demanding in a very "human" way and certainly outside the capabilities of perhaps anything short of an artificial general intelligence. I think the future of computing in the cockpit for the foreseeable future involves collaborating with humans to help them navigate these challenging situations and focus on higher level problem solving, rather than replacing them.

[semi-extrinsic]

Also, not implementing autolanding means pilots constantly practice landing airplanes. This is surely invaluable in the case of an emergency landing.