Home › View content

Computer Algorithm Beats Human Go Players

Fri, 02/23/2007 - 03:14 - NLN

The Asian board game called "Go," played by millions of people world wide, requires not only logic skills but cunning as well. It is simple in concept -- players must capture spaces by placing white and black stones on a grid in turn -- but the game requires "intuition" that is more difficult to program.
Chess algorithms are able to evaluate a move by assigning numerical values to pieces and to the strategic value of their position, something not possible for Go, where all stones are identical and scenarios are too complex. In order to play Go, computers must anticipate every move until the end of the game -- each and every time an opponent makes his move. That's why until recently even the most powerful computers have failed to gain the upper hand when it comes to playing the game against a human competitor.

Now an algorithm developed by two researchers from the Hungarian Academy of Scientists' computer lab SZTAKI helps machines compete with human Go players. So far, the algorithm can compete at the level of a professional Go player on a 9 by 9 grid. The 19 by 19 grid that top players use is still out of reach, but the new method has potential because it makes more efficient use of computer processor power than previous Go software. The more processors used, the better the new algorithm plays the game.

From Scientific American


Your rating: None

Comments

Sat, 02/24/2007 - 03:30 — CindyB

I've given it a 'go'

I played 'Go' game as a child with my father, and it really was challenging and fun. What's most interesting is that the game is simple—only two-colored stones and a playgrid—and yet the strategy involved is complex. It's an ancient game that has withstood the test of time. I had forgotten about it over the years.... Maybe it's time I broke out the old board!

Syndicate content