Reply to thread

Message: <blockquote data-quote="Fanaelialae" data-source="post: 8941974" data-attributes="member: 53980">Unless you're paying professional play testers to sit through these terrible sessions, you're looking at another problem. (And keep in mind that given the size of the datasets you need for AIs, paying play testers might very well not be a financially viable option.) So what if, for example, a sizable proportion of players downvote the "AI's" response when it (fairly) kills a character? It learns that it shouldn't kill characters? I think a lot of DMs would agree that for many tables, that should not be the take away. It's that players can be a little salty after losing a character, and one should be considerate of that. But, of course, the AI can't understand that (because it can't understand anything).At that point it has the potential to become a runaway effect. It stops killing characters, and players don't have a way to reverse this trend by upvoting when kills their character because it isn't killing characters anymore. There are some intricacies to running an RPG that are far more complex than the average conversation. Also worthy of consideration is not just whether it will be technically possible, but whether it will be economically practical to do so.</blockquote>