Explore the concept of lifelong learning in artificial intelligence through the lens of Minecraft in this 21-minute conference talk presented by Yetian Wang. Delve into the challenges of implementing lifelong learning systems and discover how Minecraft serves as an ideal testbed for developing and evaluating these techniques. Learn about skill hierarchies, DQ Networks, and Deep Skill Modules as key components in creating adaptable AI agents. Examine the application of these concepts across various domains, from simple state spaces to complex environments. Gain insights into the practical usage of skills and the significant contributions this research makes to the field of AI and machine learning.
Overview
Syllabus
Introduction
Lifelong Learning
Challenges
Minecraft
Skill Hierarchy
DQ Networks
Skills
HDRLN
Deep Skill Module
Multiple Skill Module
State
Domain
Tomb Domain
Complex Domain
Results
Usage of Skills
Conclusion
Contributions
Taught by
Pascal Poupart