Simulating Life: The Application of Generative Agents in Virtual Environments

This research explores the innovative integration of Large Language Models (LLMs) in game development, focusing on the autonomous creation, development, and governance of a virtual village by AI agents within a 2D game environment. The core of this study lies in observing and analyzing the interactions and societal development among AI agents, utilizing advanced algorithms for generative behavior modeling and dynamic skill tree learning. These AI agents are endowed with human-like decision-making capabilities, enabled by LLMs, allowing them to engage in complex social interactions and contribute to emergent societal structures within the game. The uniqueness of this project stems from its approach to simulating lifelike social dynamics in a virtual setting, thus addressing a gap in existing research and marking a significant contribution to the interdisciplinary fields of artificial intelligence and game development. By comparing AI-generated societal behaviors with human social interactions, the study delves into the potential of AI to mirror or enhance human social structures, offering a fresh perspective on the capabilities of AI in game development. This research not only aims to push the boundaries of AI applications in game development but also seeks to provide valuable insights into the potential for AI-driven simulations in studying complex social and behavioral dynamics.