Will shogi software beat male pros?
Just reading the following article now.
Will shogi software beat male pros?
Computer software has conquered one of Japan's top female shogi professionals--will top male pros be able to keep their electronic opponents at bay?
Last month, Akara 2010, a computerized shogi (Japanese chess) system, defeated Ichiyo Shimizu, the then Japan Shogi Association's women's shogi champion, marking the first time a shogi machine has beaten a professional in a public match.
Shimizu's shogi skill is said to be equivalent to male shogi professionals holding the second- or third-dan ranking, meaning the latest shogi computer program is on par with low-ranking pros.
The development of shogi software began in 1974.
Shogi players can reuse captured pieces as their own, a complication that proved too much for early versions of shogi software that could do little more than move the wooden pieces in line with shogi rules.
An experienced human shogi player anticipates future moves by narrowing down a range of potentially effective moves to several alternatives by grasping the overall situation or by making intuitive decisions.
While software programs in general take into account all possible future moves, earlier shogi software models were poor at taking stock of situations and had insufficient calculation capabilities.
Later, software developers quantified some elements needed to grasp the state of play on the shogi board and to significantly boost computer calculation performance. In 1990, it became possible to compete with--and beat--amateur shogi players ranked two- to three-kyu, the class below the dan category.
That year marked the start of annual computer shogi championships in which many software programs vie for superiority.
In chess, a computer defeated the world champion in 1997. Applications derived from chess software, along with the developers' willingness to make public software design secrets, combined to enhance the development of shogi computer programs.
Shogi software, as exemplified by Akara 2010, seems to have shed its growing pains as it already has exceeded the abilities of leading professional shogi players.
Yet shogi software still has limitations. Even though previous software versions made astute moves until the middle of a match, it tended to make bad moves especially in the end game.
A revised version of the software to address the defect was developed in just two years, using a "consensus-based decisionmaking method" employed by Akara 2010. The system has four shogi programs "consult" each other about what to do in the next move, choosing what is considered the best option based on the principle of majority rule. This reduces the possibility of making a poor move.
In the match against Shimizu played at the University of Tokyo on Oct. 11, Akara 2010 was supplanted by a "reinforcement" in addition to the four programs--a machine capable of foreseeing 60 million possible moves in one second.
Shimizu, who held the Joryu Osho title at the time, had the upper hand over the machine in the middle of the game, but she had to spend a lot time thinking through her moves. As time began to run out, Akara 2010 seized an opening and snatched victory.
According to Waseda University Prof. Takenobu Takizawa, president of the Computer Shogi Association, which has hosted the World Computer Championship every year since 1990, top male pros should be looking over the shoulders.
"Computer software might be able to defeat top male professional shogi players, even top champions such as holders of the Ryuo and Meijin titles, in five years," said Takizawa, 58.
The current Ryuo, Akira Watanabe, 26, who has played against shogi software in public, cast doubt on Takizawa's expectations.
"Ms. Shimizu had plenty of chances to win the match," he said. "I think we're still some way off from a machine that possesses the ability of a top [male] pro player."
The Japan Shogi Association has yet to decide whether to sanction another match between a top player and a computer.