Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="Clint_L" data-source="post: 8942137" data-attributes="member: 7035894">You give it information to work with. Specifically, you give it constraints. Right now, anything goes for ChatGPT doing a D&D scenario. But imagine it still has that ability to improvise, but also knows that there are limitations. So now it can respond to player input with more context. Note that it can already do this, but only if the player supplies the context. If optimized for a D&D adventure, the player wouldn't supply the context, the pre-written adventure would. Just as it does for a human DM.The question of whether it can be fun can already be answered: it can. I've used ChatGPT to do D&D scenarios and it was super fun, and very surprising. Also, folks have been using WAY more primitive AIs to have fun playing RPGs for decades.It's not all or nothing, either it's as good as a human DM or it's worthless.Edit: I'm kind of tired of arguing about what the AI can or can't do in theory, since it keeps surprising us. I suggest an experiment - I might do it if I have the time, but anyone can:Run yourself through a simple module-type solo adventure using ChatGPT. Give it the context from the module as you advance (i.e. copy and paste in the information from the adventure). Do the the dice rolling and tell it the results.My thought is to do this with the Klarg adventure from the first part of Lost Mine of Phandelver. This would obviously be better if the experiment was blind since we all have some pretty apparent biases, but give it a good faith attempt, not trying to prove a thesis but just to see what happens.</blockquote>

Verification