Prof. Dongmo Zhang
School of Computing, Engineering and Mathematics, Western Sydney University Australia
General Game Playing (GGP)
History:
IBM-DeepBlue-Watson
Trading Agent Competition (TAC)
General Game Playing (GGP)
Wiki
General game playing (GGP) is the design of artificial intelligence programs to be able to play more than one game successfully.[1]
General Game Playing is a project of the Stanford Logic Group of Stanford University, California, which aims to create a platform for general game playing. It is the most well-known effort at standardizing GGP AI, and generally seen as the standard for GGP systems.
Special Intelligence vs. General Intelligence
Human intelligence
Special Intelligence: expert at specific tasks
目标:outperform the best humans on specific task
General Intelligence: capable of taking any task
不依赖人类专家知识
目标:outperform average humans on any describable tasks
模拟人类的长期记忆、短期记忆
General Game Player
- general game player
- 可以理解Game规则: 用一种特殊语言写出来
- playing these games effectively without knowing the game rules in advance
- special:人类想怎么玩,general:计算机自己想. Goal: 0-100 & utility 也要考虑
Game description language (GDL)
- Game description language (GDL) 语形
- logical language
- role(p): p is a player
- init(F)
- true(F)
- legal(P,A)
- does(P,A)
- next(F)
- terminal
- goal(P,N)
- 目的:描述出来让计算机/人理解这个游戏
Game Model: Finite State Machine
语义
- A finite game
- Game transition
- 看走哪条路能赢
- Synchronous update: take a joint action as a move at each state
- Environment chhanges only in response to moves
- GGP Base: Standford http://www.ggp.org
Design a General Game Player
simple-player games
multi-player games
searching: Game Tree Search- minimax hueristic search(不容易找), α,β 剪枝
machine learning: monte carlo tree search
alpha go: 蒙特卡洛树搜索 【reinforcement learning】&dl
game theory ?:假设走均衡的那一步