21st Century Cup 2002 -- Bob's Commentary
For whatever reason, recent years have seen a decline in the number of computer go tournaments being held. The famous $1M prize for a computer winning against a strong human, offered by the Ing Foundation, expired in 2000, so that was also the last year that its well-attended tournament was held. The 21st Century Cup, now in its second year, ended up being one of the major tournaments of 2002.
This year's first prize was taken by "The Many Faces of Go",
a wizened veteran; its author David Fotland, the eminence grise of the
computer go scene, has been programming go for a full two decades. Many
Faces turned in a sterling 7-0 record, playing against some of the strongest
programs in the world, proving that for the time being, experience counts.
Confirming this trend was the second place (6-1) finish for Go4++,
the finely polished
program from UK-based author Dr. Michael Reiss, and the winner of last
year's 21st Century Cup. Both programs won their first three games, then
met in the fourth round. Go4++ built an early lead, but botched the handling
of a Many Faces invasion of its center territory, in the process losing
the territory, a large wall, and the game. The game record is here.
Coming in third was Go Intellect, another program with a distinguished history, from Dr. Ken Chen of the University of North Carolina. The top finisher from Asia (in fourth place) was Katsunari, a well-known program by Shinichi Sei, a Japanese author, followed by another Japanese program, Hiroshi Yamashita's Aya. Rounding out the list of six entrants finishing with four or more wins was NeuroGo from Markus Enzenberger.
In one sense, there was not much ambiguity in the results: all of the top-finishing four programs beat all lower-placing entrants.
Those seeking deeper meaning in all of this might well conclude that computer go is in a stage now where progress is coming incrementally. Potential giant-killers such as GNU Go and SmartGo turned in disappointing results, failing to win a majority of their games, although SmartGo played well against Go4++ (game record) until failing to defend properly against a simple poking move in the lower left.
This year's tournament, I think, fulfilled its objective of being a forum for authors of computer go programs to go head to head with their latest versions against the best competition in the world; to see how their programs stack up; gather real-world information about their programs' strengths and weaknesses; and to share thoughts directly with other leading authors. For observers of computer go, it succeeded in offering a fine snapshot of recent progress in the field.
|
|
|
Copyright (c) 2000, Intelligent Go Foundation. All rights reserved. |