The problem with that would be that the computers don't have the ability to recognise that the pilot might be doing something "odd" for the right reasons.
IIRC one of the Airbus crashes back when the first of the modern autopilot/fly by wire systems was in use was because the Pilot was doing an airshow manoeuvre (something that was safe) but the computer decided was wrong and the pilot couldn't over ride it.
Another near crash was averted because the pilot was able to treat a jet like a glider.
You also run into the problem that the pilot might be aware that there is something wrong with the flight computer or automatic instrumentation, but the computer disagrees with him, and anyone controlling it remotely will either have to:
Agree with the flight computer (good bye aircraft if the instrumentation is at fault), or if they're able to contact the pilot and believe him you're in the same situation as now (if the pilot is lying, good bye aircraft).
Other crashes have been caused by valves malfunctioning, but the computers didn't know that - IIRC in those instances the only thing that saved one of them was the pilots doing something that went against what the computer would have expected.
You'd also be introducing another area of possible mistakes or faults, and possibly worse, a way for the aircraft to be remotely controlled by a malicious third party who the pilot couldn't over ride.
This completely ignores the fact that if the pilot is in the aircraft they pretty much have to have access to the ability to shut down virtually any system in it, otherwise you run into the potential situation where a fire happens because the pilot couldn't shut down power when there was a short (and if they can do that, they can shut down power to the over ride system, or simply everything).
About the best way to prevent a pilot crashing the aircraft on purpose is to make sure that there is always a second person in the cockpit, fairly simple, cheap and easy to implement...