Original address: https://toyokeizai.net/articles/-/585210
Akira Hara: The true meaning that ai is a space probe by a Go player How will the relationship between AI and humans change?
Excerpt from: Toyo Economy
Author: Keiji Ume
Translation and finishing: making excuses to be quiet Content is a bit much
What kind of impact does the increasingly progressive Go AI have on modern society? In this regard, we asked the professional chess player Takumi Ohashi Seven Dan.
In 2016, a big event occurred in the Go world, and the "AlphaGo" developed by the DeepMind team of Google's Artificial Intelligence Research Institute played a five-game duel with the world's top chess player Li Sedol, and finally won 4 wins and 1 loss. Earlier, he said: "Go is a complex game, and it will take more than 10 years for AI to beat the world's top chess players", and this time AI beat the world's top chess players, which made the Go staff shake.
However, the Go world has since returned to normalcy, and after grasping the characteristics of AI, it has gradually been used to improve the strength of chess players. At present, Japan's top chess players such as Iyama Yutamoto inpa and Yili Liaoqisheng are all researching and learning according to the Go AI.
After AlphaGo, China's Tencent has also invested in the development of Go AI, and now not only professional chess players, but also ordinary chess fans can easily use Go AI.
To what extent can Go AI evolve? What does this mean for the average person? In this regard, we asked Takumi Ohashi, who has written a number of books related to Go AI.
The impact on professional chess players
- To what extent has Go AI evolved today?
Takumi Ohashi: It has become very strong, in 2016 DeepMind team defeated Lee Sedol, which had an impact on professional chess players, and then in 2017 DeepMind put in a new version of AI, defeating the Chinese chess player Ke Jie, who was considered the strongest chess player in humanity at the time.
There are a lot of Go AI now, and it's not an exaggeration to basically base it on the 2017 alphaGo version. DeepMind wrote a paper on AlphaGo, and these development companies developed it based on the content of the paper.
—— I heard that There are also differences in the level of Go AI.
Takumi Ohashi: After the retirement of AlphaGo, the "masterpiece" developed by China's Tencent should be the strongest among the current Go AI, of course, the "star array" developed by Chinese companies also has the strength to be close to the masterpiece.
There is also an open source Go AI of "KataGo", which is developed by AI developers of Go, but many Chinese enthusiasts contribute a lot of AI self-game data. In the current Go AI, Jiyi, Star Array and KataGo are the top three.
——How powerful is AI?
Takumi Ohashi: If you take the angle of making subgames, you can make more than 3 subs.
—— In the face of the highest professional nine-dan player, is the Go AI also able to make 3 children?
Takumi Ohashi: Yes, if converted into a sticker, it is almost about 30 mesh, and the gap between 30 meshes can be said to be a big gap.
——In a previous interview, Ueno Aisami Beauty Ryuki said, "I will take 3 AI to learn chess."
Takumi Ohashi: Many chess players will use a variety of AI to study chess. Many people may feel that "the answer given to us by Go AI must be the same", but sometimes different means often appear. There are also times when mistakes are made, so if you only use 1 AI to conduct research, you may miss something. Many chess players also often take a frequently used AI, and also prepare a backup AI for research.
The precision of the chess moves has been improved
With the advent of Go AI, what fundamental changes have taken place in the World of Go?
Takumi Ohashi: From the perspective of people in our time, there are some new means of Go AI, but young chess players will feel that "most of the means are actually very similar to human chess players." Of course, there are also various different opinions.
For example, the point three three this method, is often seen after the emergence of Go AI, before the mid-game stage will choose the point three three, but this method is not an innovative hand, or as if it is similar to the previous means, everyone has their own opinions in this regard.
——Rather than a huge change in chess style, it is more that the accuracy of chess has improved?
Takumi Ohashi: In essence, this is true, just now I also mentioned that Go AI can make humans more than 30 mesh, and the gap between 30 mesh is actually very large.
For example, if each person plays 100 chess pieces in a game of chess, if 1 hand loses 0.3 mesh, it will eventually lose 30 mesh. If a chess hand loses 3 or 5 meshes, professional chess players can still understand it. But a loss of 0.3 mesh in a single hand is still difficult to feel. The gap becomes very large in the end.
Because we have "accumulated 0.3 purpose accuracy", we feel that Go AI has made progress, just like a drop of water slowly becoming a river. The power of Go AI is reflected in "you lose to this opponent before you know it.".
AI seems perfect, but there are times when things go wrong.
Takumi Ohashi: That's right, Go AI sometimes makes some mistakes in the case of large-scale duels such as robbery.
After Go AI becomes very strong, the whole ethos will become "AI is right so it is right", and I am personally very worried about such a ethos. Now in the world of shogi and Go, the "winning rate" will be displayed on TV or in the online live broadcast, but this winning rate often changes greatly according to the depth of AI calculation. If ignorant people say, "What AI says is right," and it's widely known, I think it's a very dangerous thing.
Go AI is a partner in the new era
AlphaGo's deep learning technology is very ubiquitous, and it is now expected to be used in society.
Takumi Ohashi: For example, it is used in fields such as medical treatment or autonomous driving technology, but I just mentioned that these things may cause social problems.
Just like the "horizontal line effect", if everyone thinks that something is fine is repeatedly carried out, it may cause a big accident. As AI is becoming more widespread, we need to know that there will be such a possibility.
What will happen to the relationship between Go AI and humans in the future?
Takumi Ohashi: For me, Go AI is the existence of a cosmic detector. Humans can't go to Jupiter or Mars yet, but we can fly to these planets and investigate them with a space probe. In these senses, Go AI can step into a world that humans have not been able to step into.
Now That Go AI is helping humans explore new things, what Go AI hasn't seen in the future may be discovered by us humans. If we think of Go AI as a cosmic detector, I think it is our partner in the new era.
——In the current shogi world, there have been "suspected cheating" behaviors of professional chess players using Go AI, how about the current Go world?
Takumi Ohashi: The current Go world is strictly checked before or during a game. Some competitions that began in April this year also canceled their lunch breaks. Before the advent of Go AI, everyone felt that it was impossible for anyone to cheat, and some people felt that "checking the players' belongings is an offense to the players."
But now that Go AI has become normalized, there are indeed chess players who say: "In order to prove innocence, I hope that the inspection can be more stringent." In order for professional chess players to be able to play with peace of mind, it is necessary to have a perfect inspection mechanism.
Such illegal situations, I think the problem of cheating on college exams is also close to becoming a social problem. Of course, I understand some of the spiritual theories such as "such people cannot do bad things, there will be no bad people", but I think that in order to have a more sound society, I hope to complete this kind of inspection mechanism as soon as possible.
Community culture will be resurrected
As AI penetrates deep into society, will it make all kinds of human things in society disappear?
Takumi Ohashi: Actually, with the development of AI, I think some trustworthy community culture can be revived. After ai gives an international standard, it can promote communication with people in various regions. Of course, this is a medium- to long-term perspective for myself, and I think culture like the Go clubhouse will become more important.
One of the happiest things in life is a small community event with a reliable partner. It's like drinking with your friends in bars you frequent.