Imagine an AI, not just playing a game, but strategizing, deceiving, and collaborating like a human. That's the fascinating premise behind new research exploring how Large Language Models (LLMs) perform in the popular social deduction game, Among Us. Researchers created a text-based version of the game, called AMONGAGENTS, to see if LLMs could grasp the complex rules and social dynamics. Could they successfully deceive as impostors, or effectively collaborate as crewmates to identify the saboteurs? The results reveal some intriguing insights into the capabilities and limitations of today's AI. LLMs showed a surprising ability to understand the game's rules, often making decisions based on observed events. As impostors, they employed deception, strategically using vents and crafting alibis. Crewmates, on the other hand, focused on completing tasks and sharing information, exhibiting collaborative behavior. However, there is a lot of progress needed in social deduction games, as playing and mastering deception and trickery in text-based games is still a developing area for current LLMs. While LLMs demonstrated an understanding of the rules, their strategies weren't always foolproof. Impostors sometimes struggled to maintain their deception, and crewmates occasionally made inaccurate accusations. The study also explored different LLM 'personalities,' such as 'The Strategist,' 'The Skeptic,' or 'The Manipulator,' revealing how these different profiles influenced gameplay. The research highlights the growing potential of LLMs in navigating complex social situations, even within the constraints of a game. While they haven't quite mastered the art of deception, these early experiments hint at a future where AI can truly understand, and perhaps even outsmart, human social dynamics. Future research aims to refine LLM strategies for social deduction games, including more nuanced personalities and improved deception skills. The ultimate goal is to develop AI that can truly engage in complex social interactions, with potential applications far beyond the gaming world.