Researchers at Georgia Institute of Technology and Facebook AI Research have recently explored the possibility of equipping goal-driven agents with NLP capabilities so that they can speak with other characters and complete desirable actions within fantasy game environments. Their paper, pre-published on arXiv, shows that combined, these two approaches achieve remarkable results, producing game characters that speak and act in ways that are consistent with their overall motivations.
LIGHT, the platform that the researchers used to train their goal-driven conversational agent, offers a vast number of fantasy worlds containing a rich assortment of characters, locations and objects. Nonetheless, the platform itself does not set particular objectives or goals for each of the characters navigating these environments.