Transcript How computers play chess

How Computers Play Chess
Artificial Intelligence 101
Peter Barnum
November 15, 2007
“This … raises the question ‘Can a machine
play chess?’ It could fairly easily be made
to play a rather bad game. It would be bad
because chess requires intelligence.”
–Alan Turing 1946
“The decisive game of the match was Game
2…we saw something that went beyond
out wildest expectations…The machine
refused to move to a position that had a
decisive short-term advantage - showing a
very human sense of danger.”
– Garry Kasparov 1997
What move should we make?
How a computer decides
How a computer decides
How a computer decides
How a computer decides
How a computer decides
How a computer decides
Uh oh!
Adversarial search
“If I make this move, what’s the worst thing
my opponent could do?”
Examining all possible moves
…
Can I make a move that will allow me to win and
prevent my opponent from winning?
Wait, that’s easy!
35x35x35…=35N
For a game with 6 moves per player:
3512=3,379,200,000,000,000,000 possibilities
If a computer can check one billion moves per
second, it would take over 100 years
What to do?
•Can we avoid searching all possibilities?
•Can we pre-compute anything?
•Can we approximate the search?
References
• Stuart Russell and Peter Norvig Artificial
Intelligence: A Modern Approach
• Stanford Encyclopedia of Philosophy