How do humans compare to computers in board games?
In 1949, Claude Shannon published a paper that described how a theoretical A.I. system could play a reasonably good game of chess. In the past 60 years after Shannon’s paper, how far has AI come in chess performance, and more generally, in board game performance against a human opponent? Have computers unseated humans in various board games? Here are some various landmark achievements in game playing A.I.
- 1949 Claude Shannon’s Paper
- 1979 Backgammon: BKG 9.8 defeats the World Champion
- 1988 Connect Four: Solved by Victor Allis
- 1994 Checkers: Chinook takes world title against Marion Tinsley
- 1997 Chess: Deep Blue defeats world champion Garry Kasparov
- 1998 Scrabble: Maven defeats world champion and world-runner up
- 2007 Checkers: Weakly Solved
- 2008 Go: MoGo wins one of three games against a pro on a 9×9 board
Note that a full-sized Go board is 19×19, and that humans still have a considerable upper-hand over AI’s in Go. However, in just 60 years, board game AI progressed from a theoretical idea to a superhuman playing level in Backgammon, Connect Four, Checkers, Chess, and Scrabble. What advances could the next 60 years bring?