Game Playing

Question 1

Why study game playing in AI

Answer 1

Games are intelligent activities
Do not require large amounts of knowledge
Easy to measure success of failure with simple rules

Question 2

Why was Go a challenge for CS and AI

Answer 2

Larger board (5 times the size of chess board)
Not much helped by fast computation
New pieces appear every move, getting more complex
A material advantage may jest mean short term gain
High degree of pattern recognition involved in human capacity

Question 3

What is MiniMax

Answer 3

An opponent tries to prevent your win at every move
Maximise your position whilst minimising your opponents

Question 4

How does MiniMax work

Answer 4

Assume we can generate the full search tree
Game starts with computer making the first move

Question 5

How to reduce number of nodes explored in AlphaGo (2 ways)

Answer 5

Alpha-beta search based on MiniMax
Machine learning (deep learning) - supervised learning, better estimates of utility values based on human games, Reinforcement learning

Question 6

Alpha-beta pruning based on MiniMax

Answer 6

Cut off unnecessary branches for the search
Alpha - cuts at the max level
Beta - cuts at the min level