Li Shishi’s knot is not AI, but himself.
Editor’s note: This article is from the WeChat public account “ Hit the blackboard ” (ID: qiaoheiban8 ), Author Zhang Chao.
In November, a big event happened in the Go world. Former world champion and Korean chess player Li Shishi announced the end of his 24-year career and officially retired.
This decision surprised many people. After all, Li Shishi is only 36 years old, and he is still playing. Li Shishi himself explained to the South Korean media that he chose to retire because AI (artificial intelligence) was invincible.
In 2016, Li Shishi and Google Go AI program AlphaGo (Alpha Dog) engaged in man-machine warfare, losing 1 to 4, becoming the first human Go world champion defeated by Alpha Dog.
“Even if I rank first, there will be a computer that cannot be defeated. With the debut of AI in the Go game, I realized that even if I strive to become the first place, I will not be the top … … “Li Shishi said.
CNN reports that Li Shishi retired because AI “cannot be defeated”
Defeat professional players ten years in advance
The top players in the human Go world were forced to retire by an AI program. This is probably the first time in history.
Alpha Dog was developed by a team led by Google’s DeepMind Co-founder and CEO Demis Hazabis. DeepMind’s goal is to create general-purpose artificial intelligence. Hazabis proposes to start with the AI program of Go.
Demis Hassabis, Co-founder and CEO of DeepMind
Why go? Hazabis said in a public speech that the virtual environment of the game is an excellent platform for the development and testing of artificial intelligence algorithms, and Go is one of the most complex games designed by humans. Beating professional players on Go is artificial. A long and huge challenge for intelligent research.
So, the Alpha Dog is here. Its main principle is deep learning technology. The core is deep neural network. The value network is used to calculate the situation, and the strategy network is used to choose the next one. The name is AlphaGo, and there is also Google’s careful thinking behind it.
The technology behind Alpha Dogs
Alpha is the first letter of the Greek alphabet, which has the first and the beginning meaning, and go is the Japanese name for Go, because modern Go ’s system and professionalism are developed from Japan. So Google ’s AlphaGo actually has an implicit meaning of “Go first in the world”.
After the Alpha Dog was developed, DeepMind first aimed at the second stage of European Go champion Fan Wei. Born in China, Fan went to Paris after being 18 years old and won the European Go Championship in 2013.
In October 2015, Alpha Dogs and Fan Ye battled and defeated Fan Ye with a 5-0 record. This is the first time in the history of the AI program to defeat human professional Go players without letting it go, and this Originally believed to be an achievement that artificial intelligence can only achieve in 10 years.
Alfa Dog defeats Fan Yi with a 5-0 record
4 to 1 that hurt Li Shishi
Next, DeepMind chose world champion Li Shishi as the challenge.
“We chose Li Shishi because we wanted to find a legend and history player, someone who has been recognized as the strongest player in the past ten years. Choose Li Shishi to play against Go WorldThis is a major moment. Before that, the Alpha Dog had overcome all the challenges it had given it, but only after it played against the world’s top chess players like Li Shishi can everyone know the true strength of the Alpha Dog. “Hassabis once said why he chose Li Shishi.
DeepMind chooses to challenge world champion Li Shishi
Facing the challenge of DeepMind, Li Shishi also gladly took part in the competition in March 2016.
At that time, Li Shishi was standing at the top of the Go world. He won 18 world championships in important Go tournaments such as the Fujitsu Cup, Chunlan Cup, Samsung Cup, and Nongshim Cup. Top players also often have victories.
At the conference conference, Li Shishi also appeared very confident in the interview. “I don’t want to be too conceited, but I still feel that I have an advantage. Alpha Dog’s level of chess players in October (2015). It ’s not a degree to me. Considering that only a few months have passed, I do n’t think it can catch up with my level. I hope the result will be that I win 5 to 0 or 4 to 1. So my key The point is to make sure you don’t lose a game. “
Li Shishi was confident before the start of the race
However, Li Shishi guessed the score correctly, but did not guess the end result. In the end, he returned with a 1 to 4 defeat.
In the first game, Li Shishi took the lead in black. At the beginning, the format was good, but when everyone thought that the chess game was garbage time, Li Shishi made a small mistake. Alpha Dog made two strong moves and the situation reversed. In the end Li Shishi conceded.
Li Shishi is playing against Alpha Dog
Although in the second game, the Alpha Dog played a tactic that made the masters shout “Big Chess,” but the situation was already under its control. In the end, Li Shishi ran out of time and the Alpha Dog defeated Li Shishi again.
In the third game, Li Shishi tested the starting attack, life and death, and robbery. However, the Alpha Dog has always been dominant, and Li Shishi finally lost.
In the fourth game, after Li Shishi’s 78th move, the Alpha Dog began to make mistakes and produced a series of crash reactions. In the end, Li Shishi won the game. After the game, public opinion generally commented that because Li Shishi took a step in the 78th move that the computer had never thought of before, the Alpha Dog could not skip this problem and had to solve it, which caused great inconvenience to its calculation. Hazabis and Li Shishi also believe that the 78th move is equivalent to hitting the BUG of the Alpha Dog, and it is this game that makes Li Shishi the only human chess player to defeat the Alpha Dog.
Li Shishi’s 78th move in the fourth game found the “BUG” of the Alpha Dog
In the fifth game, both sides fell into a long hard battle. The Alpha Dog also entered the countdown phase for the first time. In the event of an extremely complicated situation, the Alpha Dog still won by virtue of a slight advantage.
Without Alpha Dog and Go AI
After Li Shishi lost to Alpha Dog in the first game, Ke Jie once tweeted, “Even if Alpha Dog defeated Li Shishi, it won’t win me.”
Screenshot of Ke Jie’s Weibo at that time
Ke Jie, 19 years old, was a talented chess player in China, rising from four to nine. At the age of 17, she ranked first in the world. Her record with Li Shishi was 8 wins and 2 losses.
Ke Jie ’s “challenge” getsGot a response. In May 2017, Ke Jie and Alpha Dog played at the same table.
The final result was that Ke Jie suffered a terrible defeat of Alpha Dog’s “triple kill”. In the third game, Ke Jie even cried after leaving the game for 20 minutes.
In May 2018, Ke Jie recalled the game against Alpha Dog in a speech.
“After I was led by Alpha Dog 2-0, I stayed awake all night. Because I keep thinking about how I can win it, will there be any loopholes, even if I pick a plate! I played with it in the third game! But at the early stage of layout, I made a very fatal mistake. To the middle game, I tried hard to find opportunities, but the Alpha Dog was too perfect, it played A move that made me feel cold, a move that made me feel desperate. I could no longer control my emotions, rushed out of the game room, and found a cry in an unoccupied corner. Because the upcoming 3 to 0 , This ending is too desperate for me. “
Ke Jie cries in the game against Alpha Dog
After the battle with Ke Jie, DeepMind announced that the Alpha Dog will no longer participate in the Go game, but will further explore the medical field and use artificial intelligence technology to overcome various problems in modern modern medicine.
However, the AI in Go has not disappeared because of the withdrawal of Alpha Dog. Li Shishi announced that he will also play a commemorative match after he announced his retirement. He will compete with Korean Go AI “HanDol” (Han Dou) on December 18-21 this year Play up and down three games.
HanDol is a program developed by South Korea’s NHN Entertainment. Although it was only born in 2018, it defeated South Korea’s No. 1 Shin Shin-chan in the man-machine battle on January 23 this year, and has since swept Korea. The top five Go players.
“HanDol” developer interviewed
This will be the first time that human chess players will compete with AI. The reason is that the gap between the top human chess players and AI is generally considered to be between the second and third sons. It’s just that the actual gap between human chess players and AI cannot be measured specifically because there has never been an actual match.
According to the rules of this move, AI will let Li Shishi second son in the first game. If Li Shishi loses the first game, the AI in the second game will let the third son. If Li Shishi loses in the first two games, the third round of AI will let the fourth son.
Does Go need humans to participate?
After the Alpha Dog defeated the world’s top Go masters like Li Shishi and Ke Jie in succession, some people have started to worry on the Internet. In simple areas such as Go and artistic creation, do we no longer need humans to participate?
In fact, people in related fields have already discussed similar topics. For example, in the Go world, some people in the industry have said that the professional world has almost accepted “Go AI is better than humans” and “Human chess players can go to Go “AI learning” and other concepts, even human beings can no longer represent the highest level of Go itself, has also been recognized.
Ke Jie admits that Alpha Dogs are powerful
But this does not mean that humans will not be needed in these areas in the future, or that human competition will be meaningless.
On the one hand, Go AI has discovered a lot of new moves, eliminated some human mindsets, improved human cognition of Go, and improved human skills and skills when learning from AI. Level.
On the other hand, the competition between human beings is to challenge human beings themselves, not to win over machines and programs. This is the same reason that with cars and airplanes, humans are still running.
Also threatened by AI, there is also the field of artistic creation. In 2016, a neural network named “DeepBach” created 2503 “like” Bach hymns; in 2018, an AI program of a French art team created an “Edmond de Belamy” Male portraits were sold at Christie’s for $ 432,500 (plus commissions and other fees) …
Art portrait of Edmond de Belamy by AI
However, human creativity, “creativity” and “imagination” are the most unique and important parts. Human “creativity” and “imagination” cannot be described by mathematical formulas, and therefore cannot Quantitative calculations, AI programs cannot replace human artistic creation at all until they have the “creative power” and “imagination” of human divergence.
After all, Li Shishi’s heart is not AI, but himself.
(The cover image is from Pexels)