search
search

The Rigakubu News

Disclaimer: machine translated by DeepL which may contain errors.

Shogi, AI, Research

Kohki Ikemoto, Lecturer, Department of Chemistry

 

I lost.

On Saturday, March 30, 2013, at the 2nd Shogi Den-Ou Tournament, Shinichi Sato 4-dan (currently 5-dan) gave up the game to the Shogi AI software. This was the first time a professional Shogi player was defeated by an AI player in Shogi, a symbolic battle of brains. More than six years have passed since then, and AI software has evolved to the point where it is no longer possible for humans to beat AI software. In such an era, has the existence of professional Go players lost its value at all?

Interestingly, the opposite is happening in the world of Shogi today: the new way of watching Shogi brought by AI software is increasing the attractiveness of the game between human players. Shogi is a game in which once you are outclassed, you cannot reverse the situation unless your opponent makes a mistake. The question is how to keep playing the best moves to keep the game in balance, and how to make moves that will bring chaos to the board when the game is losing ground. The AI software has made it possible for us to watch the game and see how the players' greatness and human-like qualities stand out more. This is not the only thing that AI software has brought to the table. Not only are the strategies that have been found to be effective by AI software becoming more popular, but also the strategies that were tried out long ago are being reevaluated and actively adopted. The depth of the game of Shogi itself is being reaffirmed, and we are now in an age in which we can truly feel the greatness and fun of the activity in which professionals compare readings and engage in bargaining among the vast possibilities of game play.

  The author playing Shogi. On the board, battles are being played out according to the battle plans presented as effective by the Shogi AI software.

In the research world, AI has been gaining ground in recent years. In the field of Chemistry, which is my Department of Chemistry, I have recently seen a number of papers that utilize AI. Although the use of AI is still limited to the expected range, such as "machine learning was used to optimize the condition setting" or "a new reaction was predicted and actually found as a result of machine learning," it may be possible that AI will evaluate the value of research itself and create a new sense of value in the future. Although some professors may be negative about this, it should be possible to quantify the value of research to some extent if we can construct an evaluation function that successfully quantifies the differences from previous research, novelty, and the possibilities that will expand as a result, just as Shogi AI software is now able to evaluate the form of a Shogi game. However, even if such an era comes, the importance of actually conducting research and making discoveries will never be lost, and the driving force of discovery will undoubtedly continue to be the strong conviction of the researcher.

In the countdown, we think through the problem and then make a move. When I watch a game of Shogi, I am more impressed by the attitude and way of life of the player than by the good or bad moves he or she makes. In research as well, I think we are living in an age in which the way of life of researchers is becoming more and more important, such as "Only that person can do this research.

Faculty of Science News welcomes essay submissions. We welcome all submissions, regardless of whether they are self-recommended or not. We especially welcome submissions from Faculty and graduate students. However, the Communications & Public Relations Committee will decide whether or not to publish your essay.
Please send your contributions to rigaku-news@adm.s.u-tokyo.ac.jp.

Published in the January 2020 issue of Faculty of Science News

 

Science Essays>