Interview with
summary
Minecraft runs on Java and uses modding frameworks to expose game state; Microsoft now owns Minecraft and provides an AI platform layered on those hooks. The demo implements Q-learning, follows the classic cliff-walking tutorial, and uses rewards: lava = −100, each step = −1, blue block = +100. An agent learns to avoid lava and reach the blue block, converging in roughly 35 steps on the example, showing the approach works out of the box on simple grid-like tasks inside Minecraft.
AI-written summary · about
related talks
01h 09mMinecraft as AI Playground and Laboratory
41mReinforcement Learning and Minecraft
32mAI and Minecraft
50mIEEE CoG 2019 - Industry Day - Katja Hofmann (Microsoft Research)
49mA Raspberry Pi and DeepLearning4J Brain That Plays Minecraft
36mAdding Natural Language Interface To Minecraft Using Apache NLPCraft
31mPlaying Minecraft with an AI Generated by GPT-3
56mCan I do AI?