access:
Alibaba Cloud 2022 “Cloud Procurement Season”” Hundreds of cloud products start at 0.26% off
△ Source: Wild Fox Go
What is even more surprising is that Shen Zhenzhen, who is playing white, has a 65.8% coincidence rate with AI in this game.
△ Source: Wild Fox Go
Ke Jie also exclaimed after the game:
Is this still human?
Today, this control feels stronger than that of AlphaGo, and it is impossible to play Go at all.
But it wasn’t just the loss itself that made Ke Jie shy. In fact, AI made its debut in the field of Go, and now, the game itself may have become the first batch of experimental fields for “AI to change the world“.
“A 95% win rate can be lost?”
Compared with the competition itself, the voice off the field may be the more important factor that makes Ke Jie “fatigue with AI”. Judging from his voice, he was very disgusted by “being taught by someone who doesn’t know any Go with an AI on the scalp”.
This is evident from the latest key chess game that Ke Jie lost after the battle of Shenke. In the decisive battle of the challenger of the traditional Chinese Go title tournament “Tianyuan Tournament” on March 2, Ke Jie was “turned against the wind” by Mi Yuting, who is also a post-90s chess player, and again missed the title of Tianyuan.
The winning percentage predicted by AI in this game is as follows (Ke Jie is the black first):
After Mi Yuting made a “big leg extension” that was suspected of losing the chess piece, the AI judged that Ke Jie had a huge advantage, and Black’s winning rate soared to more than 95%. But in such a situation, the final result was: in the 174th hand, Ke Jie conceded defeat.
So in the off-site comments, this voice immediately appeared:
95% win rate to lose?
Regardless of whether the advantages and disadvantages on the chessboard are really as different as AI judges, it is undeniable that in the AI era, it has become the norm to look at Go games based on the AI win rate.
Even professional commentators are not exempt. Some netizens pointed out that the idea of the current commentary is largely restrained by the winning rate predicted by AI, and there is no more imagination in the past.
And these can only be regarded as the shallow influence of AI on the ancient chess game of Go.
How “Shengong Intelligence” was made
Now, how AI has changed the game of Go may best be found in the new generation of players. For example, “Shengong Intelligence”, which led South Korea to make a comeback in the Go world.
Shen Zhenzhen, born in 2000, has a different approach to Go than other professional players. Compared with the Korean professional players who have been systematically trained in the Go Dojo since he was a child, or have apprenticeships, the way he first learned Go was through the Internet.
When he was about fourth grade, Shen Zhenzhen became obsessed with “Internet Baduk” (Internet Baduk), and sometimes even played the game of Go all night. In addition, both parents have some knowledge of Go. His father is an amateur five-dan player, and his mother is a teacher of the Go initiation class. Shen Zhenzhen quickly swept the championship in various Korean children’s competitions.
In this way, Shen Zhenzhen’s chess style is also considered to be different from others. South Korea’s “Central Daily” media once commented on the 13-year-old Shen Zhenzhen, saying that his style of learning Go through the Internet is quite “geeky”. At the end of 2016, AlphaGo, which defeated Korean nine-dan player Lee Sedol in the Go game, was officially open sourced.
Soon after, in January 2017, the Korean Chess Academy officially launched the “Korea Go Artificial Intelligence” development plan. At this time, Shen Zhenzhen also began to contact AI Go.
According to Southern People Weekly, he “spends at least 5 hours a day fighting against AI or studying chess records”. In May 2018, Shin Jin-seo defeated Lee Se-dol in the 23rd GS Cup final five-game chess tiebreaker, and in November of the same year ranked first in the Korean domestic rating.
When talking about his attitude towards AI in Go, Shen Zhenzhen believes that “although AI cannot be fully understood, its ideas can bring inspiration to him”:
Artificial intelligence has rejected almost all theories of Go so far, so we must discard existing theories of Go.
Nothing is “fixed” on the board right now, neither the midgame nor the layout is just random.
Subsequently, Shen Zhenzhen defeated famous players such as Park Tinghuan and Lee Changho in South Korea one after another. In 2021, in the semi-finals of the 22nd “Nongshim Cup”, Shen Zhenzhen defeated “the most important opponent” Ke Jie, and finally helped the South Korean team win the championship with five consecutive victories. He also began to achieve more achievements. .
Even because many Go play methods are quite AI-like, Shen Zhenzhen also has the title of “Shengong Intelligence” in the Go world.
After Ke Jie once again lost to Shen Zhenzhen, many analysts pointed out that it is precisely because the Korean Chess Academy has done a lot of research on AI that Ke Jie was completely targeted from the layout stage. Reasons for AI’s “high similarity”:
However, some netizens ruthlessly criticized Ke Jie, who lost the game, hoping that he would “balance his studies and Go training.” Others left a message “Shen Zhenzhen is still practicing chess” under the news of his returning to school + game.
In fact, the entire Korean national team, including Shen Zhenzhen, is currently using AI to assist in Go training. It is known that there are at least two Go AIs, ELF and KataGo.
Among them, ELF OpenGo is a Go AI developed by the Facebook (now Meta) team. In order to further adapt the algorithm to Go players, the Korean Chess Academy also asked four top 30 players in the world, Jin Zhixi, Shen Zhenzhen, Park Yongxun and Cui Zhehan, to cooperate with each other. Participated in testing improvement of ELF framework.
In addition, KataGo is also one of the AIs used for auxiliary training in South Korea. It was developed by Harvard AI researchers and supports dynamic stickers. By changing the number of meshes, the back game tricks are more ruthless, which is more “in line with the style of Korean players”.
But in addition to South Korea, players including China and Japan are also using AI to train Go.
For example, since 2018, the Chinese National Go Team has been using the Go AI technology developed by Tencent for intelligent training. Lin Jianchao, chairman of the China Go Association, said:
As a member of China’s national Go team “Silent”, Jueyi has well embodied the roles of coach, sparring, teacher and friend in providing winning percentage, matching degree, situation analysis and chess dismantling.
On April 18, 2019, the Japanese Chess Academy also released the “GLOBIS-AQZ” project with the goal of becoming a Go AI world champion and cultivating young players. Both the code and the project have been open sourced.
Under this trend, the Go world, which is already ruled by China, Japan and South Korea, has set off a violent AI wind.
AI changed the game of Go
There is no turning back when the bow is opened. For the Go world, the interweaving of human intelligence and artificial intelligence has become a set point. Although he has complained many times, and now he is “ruthless”, even Ke Jie himself has previously recognized the value of AI for the development of Go.
According to the China Go Association, Ke Jie once said that the “excellent art” of Go AI has provided a very crucial help for his usual training and chess progress. In addition to Ke Jie, many domestic players also agree with the view that “AI can further improve their Go skills”.
For example, after winning the Tianfu Cup world championship as a “chess veteran”, Chen Yaoye thanked Jueyi in particular, saying that it was Jueyi AI that made him improve his chess.
Not long ago, in the public “National Go Team Training Plan” for November-December 2021, the Chinese National Go Team directly stated that the players’ familiarity with AI needs to be strengthened:
Our young chess players are also doing well, but compared to Shen Zhenzhen, there is still a gap. The familiarity of AI directly affects the winning rate and time of the layout. Think about it, how did Shen Zhenzhen do it? Only by working hard and working hard can we fight against Shen Zhenzhen.
This time, after Ke Jie lost to Shen Zhenzhen in the Nongshim Cup, in addition to another discussion on the competition itself, the popularity of the AI topic will naturally not be low. For example, under Ke Jie’s news, some netizens “jokingly” recommended some books on “Artificial Intelligence” and machine learning (manual dog heads):
There is also a netizen cue to Tsinghua University where Ke Jie is currently studying, “Go to Tsinghua Artificial Intelligence Institute for help and do targeted training”:
Regardless of the perception, the trend of “AI in Go” seems to have become more and more obvious.
what you think?
Reference link:
[1]https://www.goratings.org/zh/
[2]https://mp.weixin.qq.com/s/y559XbUZ2CqJHVUMntCizQ
[3]https://mp.weixin.qq.com/s/QFwjRFuvlPZd_SJ1qfiPRg
[4]https://www.joongang.co.kr/article/10575802#home
[5]https://weiqi.qq.com/news/9435.html
[6]https://twitter.com/globis_aqz
[7]https://zh.wikipedia.org/wiki/%E7%94%B3%E7%9C%9F%E8%B0%9E