I can understand training AIs on books, and even internet forums, but I can't help but think that training an AI on lots of dumb questions with probably an excessive amount of grammar and spelling errors will somehow make it smarter.
Did they rephrase the question? Probably the first answer was wrong. Did the session end? Good chance the answer was acceptable. Did they ask follow-ups? What kind? Etc.
Or that the user just ragequit
i would consider internet forums also includes a lot of dumb questions
In 'private', people are less ashamed of their ignorance, and also know they can say gibberish and the AI will figure it out.