Sociogenesis: motivations and first steps

Sociogenesis is a project that aims to study emergent social behaviors of multi-agent systems. In its simplest form, and agent simply seeks to survive by consuming food and maximizing some happiness score. In this presentation, I will discuss some motivations and inspirations of this project, as well as the challenges to get an agent to survive on its own. I will then showcase some methods to evaluate the quality of agents and their underlying architecture beyond simply looking at their lifespan. I will then conclude with some potential applications. No prior knowledge in reinforcement/deep learning is required to attend this presentation. This work is still mostly at an exploratory stage, so discussions and inputs from the audience are more than welcome!