@nprofile1q... @nprofile1q... The point I was trying to make is that these systems are designed to give "an answer". If the question is of the type "how tall is the Eiffel tower", we can judge cleanly if the answer is correct. It's falsifiable.

If the question is of the type "why did the Roman empire fall", the answer *must* be partially wrong, no matter what it is, because that isn't a question which has "an answer". That is a "reality reference" class of question.