Title:
Towards human-like AI in Video Games

17:00 - 18:00, Dec. 16, 2022, UTC+8.

Katja Hofmann

Katja Hofmann is a Senior Principal Researcher within the Machine Intelligence theme at Microsoft Research Cambridge. She leads a team that focuses on Deep Reinforcement Learning for Games, with the mission to advance the state of the art in reinforcement learning, driven by current and future applications in video games. She and her team share the belief that games will drive a transformation of how people interact with AI technology. Her long-term goal is to develop AI systems that learn to collaborate with people, to empower their users and help solve complex real-world problems.

Abstract: 
Developing agents capable of learning complex human-like behaviors is a key goal of artificial intelligence research. Progress towards this goal has exciting potential for applications in video games, from new tools that empower game developers to realize new creative visions, to enabling new kinds of immersive player experiences. In this talk I will focus on my team's recent advances on training effective models from the rich data available in human demonstrations. The first part of my talk will focus on Uni[MASK], which provides a unified approach to offline training a single model for a wide range of tasks in sequential decision making. The second part of my talk dives deeper into the multi-modal structure that can be present in human demonstrations, and proposes a Diffusion Model for capturing these. I conclude with reflections on the open challenges along the path towards human-like AI in modern video games.