Download Outstanding Achievement @ IEEE Computational Intelligence

Outstanding Achievement @ IEEE Computational Intelligence Society
Human vs. Computer Go Competition
Chang-Shing Lee1, Mei-Hui Wang1, Olivier Teytaud2, Hani Hagras3, Hisao Ishibuchi4, and Gary G. Yen5
1
Dept. of Computer Science and Information Engineering, National University of Tainan, Taiwan
2
TAO team, Inria Saclay IDF, LRI, UMR 8623(CNRS - Universite Paris-Sud), France
3
The Computational Intelligence Centre, School of Computer Science and Electronic Engineering, University of Essex, UK
4
Dept. of Computer Science and Intelligent Systems, Osaka Prefecture University, Japan
5
School of Electrical and Computer Engineering, Oklahoma State University, USA
The game of Go originated from China. Around the 7th century, the game was imported to Japan. In the late 16th
century, the first Westerner came into contact with Go. There are millions of people that regularly play Go in many
countries around the world. Played by two players, Black and White, the stones of their color are placed consecutively
on an empty intersection of a square grid. Normally, the weaker player plays Black and starts the game. In the end, the
player who controls the most intersections of the board wins the game [1]. Go is a board game that is much more
complex than chess. Indeed, the number of possible moves in the game of Go is more important and the size of the tree
of possibilities is greater than the number of atoms in the universe [2]. However, despite several decades of artificial
intelligence or computational intelligence, there are still no computer Go programs that can challenge a strong
professional player in 19  19 games without handicap. This is because Go is a problem with high uncertainty, especially
for big board games, like the 19  19 board. Each Go player has his own way of thinking to play with his opponent, and
each top professional Go player will take different strategies even though they face the same situation. Thus, in 1997,
the IBM’s Deep Blue Supercomputer beat the World Chess Champion, Garry Kasparov, while the game of Go is still
one of the last board games where the strongest humans are still able to easily win against computers in big board games
[3, 4, 5, 7, 8, 10].
In 1998, Martin Müller won despite 29 handicap stones against Many Faces of Go, one of the top programs at that
time, in spite of 29 handicap stones, an incredibly big handicap, so big that it does not make sense for human players.
Ten years later, in 2008, MoGo and CrazyStone won Myung-Wan Kim (8th Dan Pro and winner of the 2008 US Open)
and Kaori Aoba (4th Dan Pro, 4P) in 19  19 games with handicap 9 and 7 stones, respectively [7, 10]. Since 2008, IEEE
Computational Intelligence Society (CIS), National University of Tainan (NUTN) in Taiwan, and other academic
organizations have co-hosted or co-organized several human vs. computer Go-related events, including the 2008
Computational Intelligence Forum & World 9  9 Computer Go Championship (http://go.nutn.edu.tw/) held in Sept.
2008 [3], and 2009 Invited Games for MoGo vs. Taiwan Professional Go Players (Taiwan Open 2009,
http://go.nutn.edu.tw/2009/) held in Feb. 2009 [4, 7]. In 2008, human won most of the games; however, in 2009, the Go
program MoGo made two new world records by winning a 19  19 game with 7 handicap stones against the 9P
professional Go player (Chun-Hsun Chou) and a 19  19 game with 6 handicap stones against the 1P professional Go
player (Li-Chen Chien) in Taiwan Open 2009 (http://ireport.cnn.com/docs/DOC-214010).
The flag conference of IEEE CIS, 2009 International Conference on Fuzzy System (FUZZ-IEEE 2009) held in Aug.
2009,
started
to
formally
support
the
Human
vs.
Computer
Go
Competition
(http://oase.nutn.edu.tw/FUZZ_IEEE_2009/) [5]. The games were played between four of the current best programs,
including MoGo (France), Fuego (Canada), ManyFaces of Go (USA), and Zen (Japan) against a top level professional
player (Chun-Hsun Chou, 9P) and a high-level amateur (Shen-Su Chang, 6D). During FUZZ-IEEE 2009, there was the
first win of a computer program (Fuego) against a 9P player (Chou-Hsun Chou) in 9  9 as white. On the other hand,
none of the program could win against in 19  19, in spite of the handicap 7, showing that winning with handicap 7
against a top level player is still almost impossible for computers, in spite of the win by MoGo with handicap 7 in the
Taiwan Open 2009 [7]. Also, during FUZZ-IEEE 2009, no computer program could win as black in 9  9 Go with komi
7.5 against the top professional Go player [4, 7]. The only wins in 9  9 games as black against a professional Go player
were realized by MoGo/MoGoTW against Catalin Taranu (5P) in Rennes, France in 2009 and the win against ChunHsun Chou (9P) in Taipei, Taiwan in 2009.
The human vs. computer Go competition (http://wcci2010.nutn.edu.tw/), organized by IEEE CIS, 2010 IEEE World
Congress on Computational Intelligence (IEEE WCCI 2010), IEEE CIS Emergent Technologies Technical Committee
(ETTC), and NUTN, was held in Barcelona, Spain on July 20, 2010. Several Taiwanese Go players, including ChunHsun Chou (9P), Ping-Chiang Chou (4P), Shang-Rong Tsai (6D), and Shi-Jim Yen (6D), were invited by NUTN to play
against the top four computer Go programs, including MoGo, Fuego, Many Faces of Go, and Zen, at the human vs.
computer Go competition, on July 20, 2010. A main novelty is the initial stage of 13  13 games, and the computer Go
1
programs MoGo and Many Faces of Go even won against human (6D) in 13  13 Go with handicap 2 [6, 8]. From the
games results at the competition, we know that the computer Go programs won 9 out of the total 22 games. The average
performance of the computer Go programs is fast approaching to the professional level. On the full 19  19 board,
programs have racked up a number of wins (but still a lot more losses) on 6 and 7 handicap stones against top
professional Go players [8, 9].
In 2011, hosted by IEEE CIS together with NUTN of Taiwan, INRIA team TAO of France, and Grid5000, the
human vs. computer Go competition (http://ssci2011.nutn.edu.tw/ and http://fuzzieee2011.nutn.edu.tw/go/) was held at
the 2011 IEEE Symposium Series on Computational Intelligence (IEEE SSCI 2011) in Paris, France, and FUZZ-IEEE
2011 in Taipei, Taiwan. The 9  9 blind Go games between human and computer Go programs were first held at the
IEEE SSCI 2011. MoGoTW broke a new world record by winning the first 13  13 game against the professional Go
player Ping-Chiang Chou (5P), with handicap 3 and reversed komi of 3.5. MoGoTW also won 3 out of 4 games of 9  9
blind Go and Pachi won one 19  19 game with handicap 7 against Chun-Hsun Chou (9P). Additionally, MoGoTW has
won three out of five random Go games against B. Helmstetter, ranked 4D in the Amateur World Championship in
2004. Random Go is a Go game that plays from the random initial positions, i.e. the human can choose between black
and white sides. From the results of randm Go games done with no handicap at the IEEE SSCI 2011, we can conclude
that computers are much stronger for non-standard positions when compared to humans [11].
In the three-day completion held at FUZZ-IEEE 2011, there are four programs, including MoGoTW, ManyFaces of
Go, Fuego, and Zen, invited to join this competition, and more than ten invited professional Go players accept the
challenge, including Chun-Hsun Chou (9P), Ping-Chiang Chou (5P), Joanne Missingham (5P), and Kai-Hsin Chang
(4P). The computer Go program Zen from Japan won each competition even playing 19  19 game with Chun-Hsun
Chou (9P) with handicap 6, showing that the level of computer Go programs in 19  19 game is estimated at 4D. In
addition, MoGoTW also won all of twenty 7  7 games under a specific komi, that is, setting komi 9.5 and 8.5 as
MoGoTW is white and black, respectively, suggesting that the perfect play is a draw with komi 9. The compuer Go
advances in the recent years relied on the followings : (1) simulation-based methodologies which do not rely on
evaluation functions make the Monte-Carlo Tree Search methodologies relevant in many games where such evaluation
functions do not exist , (2) good compromise between exploration and exploitation, which can be used for many other
areas far from games, e.g., planning , and (3) heavy paralllelization could be used in the clusters of multi-core machines
or grids to pre-compute the improtant parts of the games, such as the opening books [12]. In addition to Go games, the
generality of the appraoch makes it suitable for a wide application fields such as (1) energy mangement applications
[13], (2) other games, including clear breakthroughs in widely played card games [14, 15], (3) non-linear expensive
optimization, and (4) active learning [16].
Opening Ceremony of the human vs. Computer Go Competition held at
IEEE WCCI 2010
Human vs. computer Go competition held at IEEE WCCI 2010
Human vs. computer Go competition held at IEEE SSCI 2011
Human vs. computer Go competition held at FUZZ-IEEE 2011
2
In the future, IEEE WCCI 2012 will continue to hold the human vs. computer Go competition in Jun. 2012. Overall,
the events held at FUZZ-IEEE 2009, IEEE WCCI 2010, IEEE SSCI 2011, and FUZZ-IEEE 2011 were a great success
and we would like to express our heartfelt thanks to everyone who joined and watched the games. Additionally, we
would also like to sincerely thank NUTN, IEEE CIS, especially for IEEE CIS President Gary G. Yen and VicePresident for Technical Activity Committee (TAC) Hisao Ishibuchi, and the organizing committee of the conferences
who rendered kind support and great help. The authors would like to thank the National Science Council of Taiwan for
financially supporting this international cooperation research project under the grant NSC99-2923-E-024-003-MY3 and
NSC99-2911-I-024-004. Additionally, this work was supported by the French National Research Agency through
COSINUS program (project EXPLO-RA ANR-08-COSI-004). We are all enthusiastically looking forward to IEEE
WCCI 2012 (Brisbane, Australia, Jun. 10-15, 2012).
Reference
[1] E. van der Werf, “AI techniques for the game of Go,” Maastricht, The Netherlands: Datawyse b. v., 2004.
[2] EveryDayScience, Artificial intelligence: man against machine, the game of Go stand still, 2010. [Online]. Available:
http://www.scienceknowledge.org/2009/10/01/artificial-intelligence-man-against-machine-the-game-of-go-stand-still/.
[3] C. S. Lee, M. H. Wang, C Chaslot, J. B. Hoock, A. Rimmel, O. Teytaud, S. R. Tsai, S. C. Hsu, and T. P. Hong, “The
computational intelligence of MoGo revealed in Taiwan's computer Go tournaments,” IEEE Transactions on Computational
Intelligence and AI in Games, vol. 1, no. 1, pp. 73–89, Mar. 2009.
[4] C. S. Lee, M. H. Wang, T. P. Hong, G. Chaslot, J. B. Hoock, A. Rimmel, O. Teytaud, and Y. H. Kuo, “A novel ontology for
computer Go knowledge management,” in Proceeding of the 2009 IEEE International Conference on Fuzzy Systems (FUZZIEEE 2009), Jeju Island, Korea, Aug. 19-14, 2009, pp. 1056–1061.
[5] S. J. Yen, C. S. Lee, and O. Teytaud, “Human vs. computer Go competition in FUZZ-IEEE 2009,” International Computer
Games Association, vol. 32, no. 3, pp. 178–180, Sept. 2009.
[6] O. Teytaud, Computers vs. Humans in Barcelona (IEEE WCCI 2010), 2010. [Online] Available: http://teytaud.overblog.com/article-mogo-vs-humans-in-barcelona-wcci-2010-morning-54191608.html.
[7] A. Rimmel, O. Teytaud, C. S. Lee, S. J. Yen, M. H. Wang, and S. R. Tsai, “Current frontiers in computer Go,” IEEE
Transactions on Computational Intelligence and AI in Games, vol. 2, no. 4, pp. 229–238. Dec. 2010.
[8] C. S. Lee, M. H. Wang, O. Teytaud, and Y. L. Wang, “The game of Go @ IEEE WCCI 2010,” IEEE Computational Intelligence
Magazine,” vol. 5, no. 4, pp. 6–7, Nov. 2010.
[9] J. B. Hoock, C. S. Lee, A. Rimmel, F. Teytaud, M. H. Wang, and O. Teytaud, “Intelligent agents for the game of Go,” IEEE
Computational Intelligence Magazine, vol. 5, no. 4, pp. 28–42, Nov. 2010.
[10] C. S. Lee, M. Müller, and O. Teytaud, “Special Issue on Monte Carlo Techniques and Computer Go,” IEEE Transactions on
Computational Intelligence and AI in Games, vol. 2, no. 4, pp. 225–228. Dec. 2010.
[11] B. Helmstetter, C. S. Lee, M. H. Wang, F. Teytaud, O. Teytaud, S. J. Yen, “Random Positions in Go,” IEEE International
Conference on Computational Intelligence and Games (IEEE CIG 2011), Seoul, Korea, Aug. 31–Sept. 3, 2011.
[12] A. Bourki, G. Chaslot, M. Coulm, V. Danjean, H. Doghmen, J. B. Hoock, T. Herault, A. Rimmel, F. Teytaud, O. Teytaud, P.
Vayssiere, and Z. Yu, “Scalability and parallelization of Monte-Carlo Tree Search,” International Conference on Computers and
Games 2010, Kanazawa, Japan, Sept. 24–26, 2010.
[13] A. Couёtoux, J. B. Hoock, N. Sokolovska, O. eytaud, N. Bonnard, “Continuous upper confidence trees,” 5th International
Conference on Learning and Intelligent OptimizationN (LION 2011),” Rome, Italy, Jan. 17–21, 2011.
[14] C. W. Chou, O. Teytaud O, S. J. Yen, “Revisiting Monte-Carlo Tree Search on a normal form game: NoGo,” Applications of
Evolutionary Computation, Lecture Notes in Computer Science, vol. 6624/2011, pp. 73–82, 2011.
[15] O. Teytaud and S. Flory Upper, “Confidence trees with short term partial information,” 3rd European event on Bio-inspired
Algorithms in Games (EvoGames 2011), Torino, Italy, Apr. 27–29 April, 2011.
[16] P. Rolet, M. Sebag, and O. Teytaud, “Boosting active learning to optimality: a tractable Monte-Carlo, billiard-based algorithm,”
European Conference on Machine Learning (ECML 2009), Bled, Slovenia, Sept. 7–11, 2009.
3