(Original title: I'm Huang Shijie, AlphaGo human arm)
Shu Shi made from the non-Temple
Quantum Reports | Public QbitAI
Xu Fengxiong nicknamed "CB".
In 1980, Feng-Hsiung Hsu graduated from Taiwan University and was admitted to Carnegie Mellon University in 1985. In 1988, CB created the Deep Chess program. Later, after learning about Li Kaifu, he joined IBM in 1989 to continue research.
1997On May 11th of this year, Kasparov and Deep Blue (deep blue) ended their second chess man-machine battle. The human chess king eventually lost 2 to 3 to the IBM supercomputer.
This is a battle that has loaded human history.
Deep Blue is a new generation of chess computer programs developed by IBM in IBM. It is hundreds of times as deep as the previous Deep Thought.
â–³ ã„‹ ã„‹ ã„‹ å¤ è–¹è›´è‚·ç½¾ 薹蛴肷罾 薹蛴肷罾   笃 笃 笃 嘶笳
In July of the same year, Huang Shijie from Taipei’s successful high school participated in the joint examination. In the 15th issue of the Chinese language examination questions of the year, the B option was: The advantages of a modern computer are so many that it is difficult to admire.
Huang Shijie was eventually admitted to Taiwan Jiaotong University and received a bachelor degree in computer and information science four years later. In 2001, Huang Shijie was admitted to Taiwan Normal University to pursue a graduate degree.
In 2003, Huang Shijie graduated. Graduation Thesis: "The Strategy of Computer Go Robbery". After a year as a research assistant at the school, in 2004, Huang Shijie was once again admitted to the doctoral program of the Institute of Information Engineering.
In 2007On February 28th this year, students from Taiwan Normal University organized a go-to-gay party. Subsequently, this regular gathering was developed into the Shida University Chess Club. Huang Shijie was the oldest of its kind and served as its first president. In March, Huang Shijie drafted the first group table for the match. The venue was in the underground dormitory of the boys' dormitory.
â–³ çŠ ç¬ Î¥ç¼¶ Υ缶 Υ缶 Υ缶 Υ缶 Υ缶 Υ缶
In May of the same year, Huang Shijie led Shida Goqi to participate in the Taiwan Goal Cup Go tournament. There have been reports in Taiwanese media that Huang Shijie had led Shida Go Go Club to win the championship. However, after a careful lookup of the qubits, it was found that Normal University had only won fifth place in 2009, and all other years were on the list.
In the 15 individual wars of the same event, the top eight did not see Huang Shijie's name. From 2004 to 2011, Huang Shijie has been a Ph.D. student and his level of playing is amateur six.
Although he failed to achieve outstanding results in the human game of chess, Huang Shijie continued on the other route. One of his major battlefields is the computer chess program competition organized by the International Computer Game Association (ICGA). As the name suggests, computer experts from all over the world are fighting in chess, chess and other fields.
In this competition, Huang Shijie's name is written: Shih-Chieh Huang.
In 2006, Aja Go, the first Go program developed by Huang Shijie alone, won the 11th place in the Go tournament; his Chinese chess program Elephant (elephant) won the bronze medal in the tournament. In the following years, the results of Go's Go program developed by Huang Shijie have never been brilliant.
year 2010In September of this year, three others, including Demis Hassabis, set up a new company in London, England, named DeepMind.
â–³? In 010, Huang Shijie developed Erica to beat Zen to win the championship
In the same September, Huang Shijie under the guidance of Rémi Coulom developed the Goa program and beat Japan’s Zen in the Go game to win the championship. This was a great achievement at the time. Rémi Coulom is the author of Crazy Stone, another Go program.
Because Erica won the championship, Huang Shijie was invited to attend postdoctoral studies at the University of Alberta, Canada, and served as a researcher for the computer Go program. The title of his doctoral dissertation is: "A New Heuristic Algorithm for the Monte Carlo Search Algorithm Applied to Computer Go".
In June 2011, Dr. Huang Shijie passed the defense. On July 1, Huang Shijie flew to Canada.
At this time, David Silver, whom he later met, left University of Alberta a year ago and went to University College London. In London, Silva will encounter Hasabis.
After arriving at the University of Alberta, Huang Shijie continued researching Monte Carlo Tree Search. He also participated in the development of Fuego. The team also included Markus Enzenberger, Martin Müller and others. This Go participated in the ICGA competition that year, but it was nothing.
In November of that year, it was still Zen.
One year later, in November 2012, Huang Shijie also came to London to join DeepMind as a senior researcher. Starting from this time at least, Huang Shijie started using the new English name:
Aja Huang.
Early 2014During the two years of joining DeepMind, Huang Shijie did not seem to have significant research results. From the point of view of the volume of papers published, the same is true. Before 2014, he could hardly find any papers he had sent.
However, things soon turned a corner.
On January 26, 2014, Google announced a $500 million acquisition of DeepMind and won this startup that will shine in the future.
â–³ çŠ ç¬ªî€˜î’   chair
One day, Silva went to Huang Shijie and said: "Aja, we are ready to start a Go project. At first you and I only." In February 2014, the AlphaGo project was officially launched with three teams: Hasabis, Silva, and Huang Shijie. Hasabis is the owner of the entire company and Silva is the manager of Huang Shijie. Therefore, only Huang Shijie is the only one who really works in this team.
The AlphaGo project is trying to come up with a powerful Go program. And from the very beginning, the team decided not to try all the methods. They had only one direction: exploring in the direction of deep learning and reinforcement learning. It was also from the very beginning that they knew it was a very difficult challenge.
Hasabis said that he hopes to make the machine gain intuition and creativity through AlphaGo's research.
The more realistic difficulty is that compared to chess, Go has a huge computing space, and the computer cannot understand who wins a game.
But they set off like this.
June 2014 - June 2015This summer, AlphaGo's first important weapon appeared. With the help of a convolutional neural network, AlphaGo learned a lot of human master's chess and could make a human instinct in 3 milliseconds. Later, Huang Shijie named the weapon: "Strategic Network," and continued training optimization.
At this time, AlphaGo's training was still done on the GPU.
One year later. In June 2015, AlphaGo had more powerful and distributed search technology support. The ability to read the game is greatly enhanced, allowing you to retrieve a variety of changes and find the best way to deal with it. At this time AlphaGo can calculate the follow-up 40-60 moves.
(Qubeit Note: The defeat of Ke Jie's latest version of AlphaGo is only counted as 50 moves.)
At the same time, under the same hardware conditions, AlphaGo achieved a 70% win rate for another Go program Crazy Stone. In other words, the chess force led the way. This achievement has encouraged the entire DeepMind. Hassabis told Huang Shijie at this time:
"Aja, let's form a team. You don't have to go it alone."
From this time on, more and more deep learning engineering experts have gradually joined the AlphaGo team. Huang Shijie also organized a training class for newly joined colleagues to popularize the basic rules of Go.
Delta Card
August 2015At the same time, another important node is not far away. Two months later, AlphaGo will master the most critical ability to dominate the world of chess: situational judgment.
“Actually, situation judgment is the most difficult and most vexing aspect of the Go process. To make accurate judgments, we must have the ability to accurately measure the size of both sides, and at the same time, we must have a comprehensive view of the general situation and tap potential value. Analytic ability and reasoning ability... To make a truly horizontal game, situation judgment is very necessary.â€
The above passage comes from Li Changyi. The South Korean player, nicknamed "Shifo", began his first world championship in 1992 and has won 18 individual championships and 13 team championships since 2007, creating the "Li Changjun era."
How does AlphaGo get the ability to judge the situation?
The person who solves this problem is Silva. One day he said to Huang Shijie: "Aja, I have an idea. I think it might work." Silva’s idea was later called "value network." At that time, Huang Shijie was very skeptical of this idea. He replied, "Can you use it? Let's try it."
The value network is also a convolutional neural network. The input is the position of the roll, and the number between 0 and 1 is output. 0 represents the victory of the opponent, and 1 represents the victory of the opponent. If it is almost the same, it outputs 0.5. (Qubit Note: Google’s recently released figures are -1 to 1, slightly different).
In short, AlphaGo trains the value network through self-fighting, and then can judge the winning percentage of each hand. This forms the ability to judge the situation.
The emergence of a value network made AlphaGo's chess power soar. In the match with Crazy Stone, AlphaGo's winning percentage reached 95%, which is to reach the level of two sons.
"At the time AlphaGo had been able to crush me. I already felt its power," Huang Shijie recalls.
October 2015Also in August, Fan Hao attended the European Go Games and returned to his home in France. He received an email from Maddy on the AlphaGo team. The content of the email is very simple: We are a London company and we hope to invite you to our company. There is no more information.
Fan Li once thought it was spam. But the ghost makes a difference. He returned an e-mail and said, "Yes." Then they used Skype to communicate online. Silva also participated at the time. They told Fan Yi about the interesting project being done and also told him that he was a Google-acquired company.
Afterwards Fan Li went to DeepMind's official website, and it was determined that there was such a company and the company’s home page also mentioned Go. So he decided to go and see it.
The first time the two sides met was at the end of September. When he decided that he was going to play against a Go program, the whole person relaxed and said, "For a piece of software, it's not a matter of minutes." At the time, Fan Xi even told the AlphaGo team that he only needed an hour to play with AlphaGo.
At that time, Huang Shijie repeatedly told Fan Xin that AlphaGo was very powerful. However, Fan Yan could not hear it.
â–³ çŠ è©ˆç¬ ç—ª 馕 馕 氖蟀 氖蟀 氖蟀
October 5-9, 2015. Fan Li went to London again and had five rounds with AlphaGo. At the time, Huang Shijie was the person who had replaced AlphaGo. As a result, we may all know that Fan Yi lost all. "At that time, my entire Go world collapsed." Fan Yan said.
DeepMind was not released until the beginning of 2016. At that time, Fan Hao had been hired to become AlphaGo's coach. On the day of the announcement of the results, Fan Hao closed the cell phone. Fan’s wife, who had gone out to buy food, called the home phone: “Don’t read the comments on the Internet. It’s hard to hear.â€
Nie Weiping commented at the time: “Fan Shu’s level is too low and it gives us shame.â€
2016On January 27th, "Nature" magazine introduced the AlphaGo developed by the DeepMind team in the form of a cover story, and it defeated the European champion Fan Yi.
Silva and Huang Shijie are the first authors of this paper.
After defeating Fan Li, AlphaGo's value network, strategy network, and search capabilities have continued to increase. The hardware used has also changed from GPU to TPU. TPU has greatly improved AlphaGo's computing power.
From March 9th to 15th, AlphaGo and Li Shijun battled five games. AlphaGo eventually won 4:1. At that time, sitting in front of Li Shizhen, Huang Shijie was replaced by AlphaGo.
â–³ çŠ ç¬ çŠ ç¬ î€š  埽 埽 渥 渥 渥 æ°– æ°– æ°– æ°–
Huang Shijie first appeared in the sight of the global audience.
Huang Shijie, like a robot, appeared in the eyes of Li Shizhen and the global audience.
Li Shijun later recalled: “Huang Shijie is the hardest person involved in this man-machine match. He fears that I will be affected. During the game, he has been expressionless and has not even been to a bathroom.â€
A game can take up to six hours. In the game with Li Shizhen, Huang Shijie only drank a sip of water. Huang Shijie's performance even went into the opponent's dream.
After the hotel’s breakfast was delivered, Li Shi’s wife awakened him and had breakfast. Unexpectedly, Li Shih-hsang actually replied, "Well, I want to eat with Aja."
After the match with Li Shizhen, Huang Shijie had a short vacation and had almost no report coverage left. A Chinese reporter named Fred Zhou told the qubit that during the Korean game, Huang Shijie was not allowed to be interviewed, allegedly because he was accustomed to expression without reservation.
Then DeepMind released the game and AlphaGo continued to grow. On November 18 of this year, Huang Shijie concluded: "One of my latest experiences is that people's progress is used at best, and computer progress is used to fly."
At the end of the year, AlphaGo is back. On December 30, 2016, Science News released a tweet that has nothing to do with the content. The tweets are as follows:
AlphaGo: "Now, I am the master."
2017From the evening of December 29th, 2016, a user named "Master" marked himself as a nine-segment player of Korea. He continuously killed four players in the game platform of Yicheng and Yehuo Fox and successively "killed" all the players of Go. , Including: Ke Jie, Pu Tingyi, Jingshan Yutai, Chen Yaoqi, Shen Zhenwei, Chang Hao, Coulee, Zhou Ruiyang...
The battle that the Master provoked ended in a 60-game winning streak. Ke Jie lost three games in a row and was hospitalized for the first time because of acute gastroenteritis. This experience was also compiled into a small piece of music.
After winning the 59th consecutive victory, Master announced his identity in the chat room: "I am Dr. Huang of AlphaGo." Sure enough, AlphaGo was really Huang Shijie.
â–³ 牷 牷 èŒ èŒ èŒ
“I very much hope that I will have the opportunity to explore Go together with the Go organizations and experts this year, and to jointly comprehend the mysteries of Go in an atmosphere of mutual enlightenment. We hope to announce other announcements as soon as possible,†Hassabis said in an official statement later. Again hinted at the upcoming battle between AlphaGo and Ke Jie.
At 3 o'clock in the afternoon on April 10th, at the press conference of the Chinese Chess Academy, Ke Jie and AlphaGo were formally announced to open the man-machine war in May. Ke Jie's position at the time would be to defeat AlphaGo by any means: "I will not say Losing it does not matter, I have a conviction that must win and a faith that must die."
On May 23, from early morning to evening, the rain was held as scheduled. Ke Jie is opposite, sitting still is Huang Shijie, still serving as AlphaGo's human flesh arm.
Next to Ke Jie's table, apart from a cup, there are two plates of fruit and snacks. On the table on the side of Huang Shijie, there is only a white porcelain cup. This is true in all three games.
â–³ banana oyster
After the end of each game, Huang Shijie will quickly return to the on-site viewing hall from the opposing room. However, he just stood silently in a corner, looked down at his mobile phone, and sometimes looked toward the stage, waiting for Ke Jie and others to stand in the spotlight. He was not tall and he would often be blocked by people around him.
After the second match ended, Ke Jie talked about Huang Shijie on the stage:
"Dr. Huang is a very important person in AlphaGo. I think he is sitting in front of me like a robot (laughs). If Coulee is sitting opposite me, he may see that AlphaGo's method may be surprising Look."
"Dr. Huang's attempt to play chess may make me feel more like AI. I also admire Dr. Huang. When we watch live, we can see that I am very moving and always like to move around. Dr. Huang is always motionless. Do not go to the toilet, drink water or eat anything... so when Dr. Huang was playing chess, I think Dr. Huang is AlphaGo.â€
At the time, Hassabis also said: "Dr. Huang is incredible. Sometimes we also think he is a robot. I think he must have practiced in the mirror before."
May 27, 2017"The last game is about to begin. I have to play AlphaGo's every move." At 9:35am on the last day of the man-machine war, Huang Shijie made such a statement in a circle of friends.
The qubits had a strange feeling at the time. One is that Huang Shijie has never made a sound before the game, and the second is that there is a clear and strong feeling in this statement.
In the final battle, Ke Jie was as emotional.
At 1:06 that afternoon, Ke Jie took off his glasses and looked depressed. Two minutes later, Ke Jie got up and left. Instead of AlphaGo's 127th Dr. Huang Shijie, he quickly picked up the glass and drank. Then Huang Shijie resumed his calm expression and looked down at the board.
After a long time, Ke Jie had not returned yet. Huang Shijie also began to look up. After more than ten minutes, Ke Jie finally returned. Later we knew that Ke Jie cried after hiding behind the site's propaganda board.
At 1:22 pm, Ke Jie wiped his eyes again. At this point, the two sides reached the 129th hand.
After 209 hand-to-hand encounters, Ke Jie then lost AlphaGo.
"In fact, when I was playing today, I was a little out of shape, because it's really perfect under it (AlphaGo)... I can only guess it's half the game. This is a huge gap between me and it." After Ke Jie sobbed on the stage to tell the feelings of the last battle.
Not later on, Hasabis, who was on stage with him, said that AlphaGo has been "retired" since then. Only then did the qubits understand the meaning of Huang Shijie's morning friend circle.
When all of the above occurred, Huang Shijie sat in the first row under the table, in the middle of the position, listening quietly with his head up. I couldn't see what he was doing and couldn't see any action. There is still no opportunity to speak on stage. With the media close, Huang Shijie is still saying: They will not allow me to interview.
At 0:34 in the morning on May 28th, Huang Shijie issued a sentence in his circle of friends: “There is no such thing as a feast for the rest of the world. The time has come for me to transfer to other projects.â€
All this ended.
This is another battle that has loaded human history.
At last“In recent years, especially after coming to work in the UK, I sometimes feel that life is like dreaming. In the team and my colleagues, I also saw the true meaning of people because of their dreams.†A colleague asked Huang Shijie early this year. Did you ever think of today when studying Computer Go? He wrote this passage.
From the beginning to the present, Huang Shijie looked at AlphaGo's birth, growth, and fame. Huang Shijie has been appearing as an AlphaGo human arm, like a robot.
But Huang Shijie is far more than the human arm of AlphaGo.
He is also not an indifferent robot.
Interaction time
During the man-machine war, the qubits were signed by several core figures. As shown in the figure below, can you recognize these signatures and who are you?
ZGAR Vape Device 5.0
ZGAR electronic cigarette uses high-tech R&D, food grade disposable pod device and high-quality raw material. All package designs are Original IP. Our designer team is from Hong Kong. We have very high requirements for product quality, flavors taste and packaging design. The E-liquid is imported, materials are food grade, and assembly plant is medical-grade dust-free workshops.
From production to packaging, the whole system of tracking, efficient and orderly process, achieving daily efficient output. We pay attention to the details of each process control. The first class dust-free production workshop has passed the GMP food and drug production standard certification, ensuring quality and safety. We choose the products with a traceability system, which can not only effectively track and trace all kinds of data, but also ensure good product quality.
We offer best price, high quality Vape Device, E-Cigarette Vape Pen, Disposable Device Vape,Vape Pen Atomizer, Electronic cigarette to all over the world.
Much Better Vaping Experience!
E-Cigarette Vape Pen,Disposable Device Vape,Vape Pen Atomizer,Latest Disposable E-Cigarette OEM vape pen,OEM electronic cigarette
Shenzhen WeiKa Technology Co.,Ltd. , https://www.oemvape-pen.com