| BackusMcStabb |
I recently started a Curse of the Crimson Throne campaign as a test for using ChatGPT as a DM. I rolled up 4 characters, used Astra to create character sheets with all the relevant information, and uploaded the finished sheets to a project. I uploaded the first CotCT book, map folio, GM instructions for Chat, a party reference journal, a Campaign State document, and a Campaign journal.
I did all the rolls myself and gave the results to Chat to add the modifiers and basically just had it play NPCs and enemies and follow along through the book.
Character creation with Astra went incredibly well. No errors, it caught things I missed and made great suggestions. The sheets came out perfectly.
Session One and Two also went very well. Session 2 went totally off script and resolved one of my character’s backstories before getting back to the book in session 3. Session 3 began to have some issues. NPCs got confused; they knew things they shouldn’t have and injected themselves into previous experiences that they hadn’t been a part of. It also made some errors when I upgraded gear; calculated gold wrong and didn’t correctly update bonuses. When I went to session 4, it seems to have reached a point of critical failure. It picked back up towards the end of session 3 but it didn’t account for the things that happened last.
I’m curious if anyone else has used AI in a similar way. What did you find worked or didn’t work? Any tips and tricks to fixing it‘s continuity capabilities?
I should note that while I had Astra create and update character sheets, I had to use Sol for the actual gameplay. Too credit intensive otherwise.