@jeffreyrb03-gmail This is an interesting approach to creating AI enemies in a game! I can see how using ChatGPT blocks could add some unpredictability to enemy behavior. The letter-code system is clever for simplifying input/output.
That said, I wonder about the response time issues you mentioned. Even a couple seconds delay could feel frustrating in many game types. The other challenge might be maintaining context over a long game session - LLMs sometimes “forget” earlier parts of conversations when they get lengthy.
I’ve been experimenting with something similar while using narrati.io for my game narrative development. While narrati is primarily for generating story content, I noticed its contextual memory could potentially be adapted for simple game AI behavior patterns too. I was drafting character personalities and reaction patterns, and realized these could translate to behavior rules for NPCs.
Maybe a hybrid approach could work? Use pre-defined behavior patterns (developed through story-oriented tools like narrati for quick reactions, but supplement with ChatGPT for more complex decision-making during natural pauses in gameplay? This way you get the best of both - responsive gameplay plus intelligent adaptation.
Would be curious to see what you end up building with this approach! The concept definitely has potential for creating more dynamic enemies than traditional hardcoded behavior trees.

@jeffreyrb03-gmail This is an interesting approach to creating AI enemies in a game! I can see how using ChatGPT blocks could add some unpredictability to enemy behavior. The letter-code system is clever for simplifying input/output.
That said, I wonder about the response time issues you mentioned. Even a couple seconds delay could feel frustrating in many game types. The other challenge might be maintaining context over a long game session - LLMs sometimes “forget” earlier parts of conversations when they get lengthy.
I’ve been experimenting with something similar while using narrati.io for my game narrative development. While narrati is primarily for generating story content, I noticed its contextual memory could potentially be adapted for simple game AI behavior patterns too. I was drafting character personalities and reaction patterns, and realized these could translate to behavior rules for NPCs.
Maybe a hybrid approach could work? Use pre-defined behavior patterns (developed through story-oriented tools like narrati for quick reactions, but supplement with ChatGPT for more complex decision-making during natural pauses in gameplay? This way you get the best of both - responsive gameplay plus intelligent adaptation.
Would be curious to see what you end up building with this approach! The concept definitely has potential for creating more dynamic enemies than traditional hardcoded behavior trees.