The 'light' variant of this question makes a lot of sence, though:

The askers were non-technical but very jaded in human nature, so they know the human-provided input might be slightly imperfect but not completely wrong.

Then the question is: Can a machine based on perfection cope? Will it output reasonable but slightly-wrong conclusions? Will it go off the rails and produce completely insane errors?

Answering this question requires a lot of interesting subtility, which has been completely rejected by babbage.