Overview
Explore the potential of Minecraft as a platform for advancing artificial intelligence research in this captivating conference talk from CHI PLAY '19. Delve into Project Malmo, an open-source experimentation platform built on Minecraft, and discover how it enables a wide range of AI research opportunities. Learn about the MineRL competition, designed to drive advances in sample-efficient reinforcement learning with human priors. Examine the challenges and opportunities presented by Minecraft's open-ended nature, including complex decision-making and collaboration. Gain insights into current AI approaches, sample-efficient learning techniques, and the potential impact on video game development. Uncover the exciting possibilities at the intersection of AI and gaming, and explore how these advancements could shape future game experiences.
Syllabus
Introduction
My day job
Why Minecraft
Reinforcement Learning
Reinforcement Learning Agent
QQ Learning
Anthony Diggle
Fraser Ellison
Diego Parinibbana
Benchmark Tests
Current Solutions
Current Challenges
Benchmark Results
Generalization Techniques
Results
No Regularization
Mineral Competition
Mining Competition
Summary
Resources
Learning Problem
Taught by
ACM SIGCHI