Autotelic Agent

Autotelic agents are artificial intelligence systems designed to learn and develop skills autonomously, driven by internally generated goals rather than external rewards. Current research focuses on enhancing their ability to generate diverse and challenging goals, often leveraging large language models to represent and manipulate these goals, and employing graph neural networks to improve skill representation and transfer. This research aims to create more robust and adaptable AI systems, potentially impacting fields like robotics and education by enabling the development of agents capable of mastering complex tasks and continuously expanding their capabilities.