Original address: https://www.tokyo-np.co.jp/article/158220
Akira Hara: [Full version] Title holder talk that Sota Fujii of shogi and Kotaro Seki of Go talk all the time
From: Tokyo Shimbun
Author: Takataka Morimoto
Translation and finishing: Make excuses for being quiet
At present, shogi shogi Satoshi is crowned with four crowns, and Seki Kaitaro Tengen, who has won the title for the first time. Two young men who won the titles of the Battle of the Throne and the Battle of Tengen sponsored by the Tokyo Shimbun News Agency talked together as they wished. Both players love to use artificial intelligence, how to face the chess game, how to use AI, and the relationship with the master and apprentice, from all aspects to develop a variety of topics. This time, we shared the contents of the newspaper that could not be published to everyone without reservation.
Satota Fujii was born in Aichi Prefecture in 2002. In 2016, he became a professional chess player at the youngest time in history at the age of 14 years and 2 months. In 2020, he won the title of chess saint, becoming the youngest chess player to win the title, and has since won the title of throne. During the conversation, Satoshi Fujii also held four crowns of the Dragon King and the King of The Dragon.
Kankai Taro was born in Tokyo in 2001. In 2013, he became the first Japanese player to win the championship at the World Junior Go Championship. He became a professional chess player in 2017 and won the title of Tianyuan in 2021. Winning the title four years and eight months after the final period is the fastest in the history of Japanese Go.
Satora Fujii: Go is just eating melon masses
- Evaluate each other's performance.
Satota Fujii: Kwan Hang Taro defeated IchiriKudan in the Battle of Tengen, and felt that his strength increased very quickly.
Kwan Hang Taro: I almost won the title before I was 20 years old, but Satoshi Fujii has already won four crowns by this time. Watching the news, I think this one is very powerful.
——What about Shogi Shogi by Kane Kaitaro and Satoru Fujii at Go?
Kwan Hang Taro: I basically don't know shogi, I only know the rules.
Satota Fujii: I don't know anything about Go, but double-alive and hijacking are ok (Note: Shuanghuo and Hithiking are in Japanese as セキ and コウ, respectively, and Kwan Hang Taro's affectionate name is "セキコウ"). Watching the live broadcast I couldn't understand at all, but it was very interesting.
Kwan Hang Taro: Watching the live broadcast of shogi, I was shocked to see them play late.
——In addition to their respective professional fields, do they have their own areas of concern?
Kwan Hang Taro: There have been a lot of matches lately, and I don't have time to think about it except for Go, so since this is a rare opportunity like this, I want to try shogi.
Satora Fujii: Recently I was practicing Chess Problem. Chess and shogi have a lot in common, but there is still a little difference, which is very interesting. If I play Go, I will just be a melon eater.
Kwan Hang Taro: Watching AI play chess can learn things
How do you use AI?
Satora Fujii: I started using shogi software when I was in the third dan of the reward meeting, and I used AI to study the chess games I had played, and sometimes I played against AI.
Kwan Hang Taro: In 2017, AlphaGo defeated the world's strongest Ke Jie Jiudan, giving people a great impact. After a year or two of fixing, I started researching with AI. At that time, the chess players around me had not yet started to use AI, so I was ahead of everyone in this regard.
It is said that Kwan Hang Taro often watches the matchup between AI.
Kwan Hang Taro: I think I already like to do this kind of thing rather than learning chess, but there have been a lot of games lately, and the time has become relatively shorter. In the middle stage, human words will play the means within common sense, but AI will make the means that we can't imagine. The most important thing is that there are almost no mistakes in the official stage, so I often learn from the layout to the end.
—— Can this be different from watching the game between human chess players?
Kwan Hang Taro: If you look at it carefully, you will feel that you can learn more from the AI left-right competition. However, many places have not been able to understand thoroughly, looking at it is very tired, this time to see the human chess.
Satoshi Fujii: I sometimes watch the matches between shogi software, but it would be very laborious for me to think about the intention of each hand, so I just watched and played.
New discoveries have been made after learning about deep learning
At the end of 2018, not only the Go community, but also the "AlphaZero" chess score that can play shogi and chess was made public and became a topic. Did Satora Fujii read the chess game?
Satoshi Fujii: Yes, the shogi software that used deep learning technology at that time was far less powerful than it is now, and compared to the software of that year, AlphaZero gave me the impression of a lot of unique methods and judgments. However, I can't refer to the chessboards that are public. Later, the "dlshogi" that used deep learning technology was very powerful, and after using this software, I found a lot of new things, which was very interesting.
When we interviewed Hassabis that year, he said, "Go is a very elegant game." Do the two of you feel the beauty of chess in their respective fields?
Satora Fujii: After all, there are games to win or lose, so it is difficult to feel the beauty directly. In terms of aesthetics, shogi is a condensation of the content of aesthetics, and there is indeed a little beauty in this regard. However, shogi also has accidental results, and it is common to play under such a situation. This approach is also impressive.
Kwan Hang Taro: To be honest, I don't feel how much beauty there is in human chess, I think that when humans and AI are playing against each other, AI's chess is really very beautiful.
AI "Wins Firmly in the End"
——When Kwan Hang Taro looks at the layout, can he already feel how much percentage THE AI will give?
Kwan Hang Taro: The Go board is very large, even if the situation is the same, the winning rate will be different. So it takes a lot of time in the layout phase, and although there will be a little error, you will probably know what the winning percentage is when you start the layout phase to the middle of the game.
Satota Fujii: As long as you end the trail, it is difficult to judge the situation accurately, and I think it is very powerful to be able to do this.
——After the emergence of AI, I feel that what has changed, what place is still the same.
Kwan Hang Taro: I feel that it is easier to learn chess. However, the essence of my chess and the style of chess have not changed, and we use AI to study it in the layout stage, but in the middle game, my chess is biased towards human play.
Satota Fujii: I still benefited a lot for myself, before the advent of AI, I never understood how the layout was evaluated, but after using AI, I gradually understood which chess types were good players and which were bad hands. It's really hard for us to use AI in the mid-market stage, so overall it doesn't change much. In the final stage, it used to be repeated, but after reading the evaluation of AI, it was found that AI often won steadily in the end. Realizing that there was still this way of winning chess inspired me a lot.