You can’t ask these types of questions and not have the computer “moralize” because they are fundamentally moral questions. You and I have no problem saying hi

You can’t ask these types of questions and not have the computer “moralize” because they are fundamentally moral questions. You and I have no problem saying hitler was worse but a Nazi party member from 1940 would likely say Hitler was obviously better because we have different moralities. Questions with explicit facts like “show me the founding fathers of the United States” that were all known actual people showing up with wildly different looks is one failure mode of these systems but I keep seeing commentators in this post bring up questions that do not have a “correct” answer without looking through a specific moral viewpoint, and getting bent out of shape that the model is responding with an answer through their own personal lens Edit: I should read a whole comment before responding, I just restated what you meant. Think it’s time for more coffee