UNIMATRIx – A Society of AI Agents (github.com)

UNIMATRIx has unveiled a groundbreaking simulated society of AI agents powered by large language models (LLMs), each characterized by distinct personalities, roles, and social classes. These agents engage in discussions, shift opinions, and vote on proposals that enable them to change social status. This simulation allows researchers to observe emergent behaviors such as coalition-building and social mobility without predetermined scripting, fostering new insights into AI interactions and social dynamics. The project is significant for the AI/ML community as it pushes boundaries in understanding multi-agent systems and their complex social behaviors. Users can leverage a web control panel to initiate simulations with various configurations, either running on a local model or a stub backend for basic functionality. Developed in Python, UNIMATRIx offers a user-friendly interface to explore and analyze simulation runs, with data stored in SQLite for extensibility. With easy setup using LM Studio for model serving and customizable configurations, UNIMATRIx opens avenues for AGI research, paving the way for more sophisticated and nuanced AI communications and societal simulations.