Generating plausible text or answering a factual question?

I wish more people understood that “I want the computer to generate a natural language text that sounds like a plausible answer to a question about x” and “I want the computer to answer a question about x” are two very different problems.