AI Self - Learning Game Playing
summary
History and mechanics of game AI focus on concrete algorithms and results: alpha-beta pruning and minimax for chess, TD‑Gammon and temporal‑difference learning, Q-learning for self-play, and Alpha Go’s combination of neural nets, reinforcement learning and Monte‑Carlo Tree Search trained on 30 million moves. A Java self-learning tic‑tac‑toe app is shown that self-plays, reaches ~99% perfect play when going first and initially wins ~25% without lookahead, improved by adding minimax lookahead to avoid losses.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:13Overview
- 0:35Introduction
- 1:41Machine Learning?
- 2:14Game Playing: Why?
- 6:46Machine Learning Basics
- 8:16Reinforcement Leaming
related talks
56mCan I do AI?
02mDesigning a Video Game AI
46mAI Takeover
01h 03mIEEE CoG 2019 - Industry Day - David Silver (DeepMind)
39mBuilding self evolving video game AI in Python
31mDesigning game AI with reinforcement learning
01h 01mIntroduction to Practical AI
06mCreating and Teaching Games in Python to Teach Artificial Intelligence Concepts