A View from Michael Patrick Gibson
The Games Computers Play
With 18 years of data, a checkers-playing computer program knows every possible winning move.
A team of computer scientists at the University of Alberta has created a computer program that’s incapable of losing a game of checkers. An opponent’s only hope is for a draw. To create the system, the team spent 18 years processing every possible move–in this case, 500,000,000,000,000,000,000 checkers positions. If your eyes got lost in all those zeroes, that’s 500 billion billion. Such an effort is known as a brute-force method because it entails using an algorithm to excavate a definite answer–is this a winning move?–out of billions and billions of possible outcomes deep in a decision tree. An earlier version of the checkers program, which won the World Man-Machine Championship in 1994, used a different style based on probabilistic inferences about winning strategies.
It has been slightly more than 10 years since Deep Blue beat Gary Kasparov in chess using a brute-force technique similar to the Alberta team’s Chinook program. And since then, there has been much debate as to whether these types of programs are intelligent and, if so, whether we are a less unique and therefore less precious species than we believe ourselves to be. But we have long known that a calculator can do a sum faster than any human, so why should we feel threatened by any game-playing software?
In 1814, the mathematician Pierre-Simon Laplace hypothesized the existence of a supercomputer capable of using a brute-force method to think through all the possible movements within the universe, from the smallest all the way up to the largest bodies. For such a computer–it became known as a demon–Laplace said, “Nothing would be uncertain and the future just like the past would be present before its eyes.”
Could it be that some feel uneasy about losing games to artificial intelligence not because they have to share what they believe to be a uniquely human capacity but, rather, because each adversarial thinking machine is potentially a Laplace mini-demon?
While you ponder that, why not play, and probably lose, a game of checkers with Chinook?
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today