AI and Minecraft Lars
summary
Uses Project Malmo to run agents inside Minecraft via mission XML and a Python API, with hooks to Lua/Torch and C#/Java clients. Shows a scripted demo that finds a diamond block, then a Q-learning example that learns cliff-walking with rewards: -1 per step, -100 for lava, +100 for the blue goal, +500 for a distant gold goal. Demonstrates a Q-table that converges over repeats and contrasts image-based deep reinforcement learning (Atari/Breakout at ~100/20/400 training steps). Notes the learned policy fails to generalize to new maps.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:46Minecraft
- 2:57Minecraft demo
- 5:32Project Mountain
- 7:41Example
- 10:27Supervised Learning
- 11:28Reinforcement Learning
- 12:04Trial and Error Learning
- 12:23Cliff Climbing
- 12:51Cliff Walking
- 17:23Blue Block
- 21:13Yellow Block
- 22:33Arcade Learning
- 25:10Conclusion
related talks
32mAI and Minecraft
41mReinforcement Learning and Minecraft
01h 09mMinecraft as AI Playground and Laboratory
50mIEEE CoG 2019 - Industry Day - Katja Hofmann (Microsoft Research)
41mLearn AI automation with Minecraft
56mCan I do AI?
49mA Raspberry Pi and DeepLearning4J Brain That Plays Minecraft
26mChatGPT and Minecraft's diamonds playing agents: Reinforcement Learning in a nutshell with sheeprl