The man-machine battle is about to begin. AlphaGo vs. Lee Sedol. Who do you bet on to win?

In March, a major event that will attract worldwide attention will take place in the technology world.

Apple releases iPhone 5se? No no no, this is just an industry event. Anyone who pays a little attention to the latest technology news should have guessed that this "big event" is the match between Google's Go system AlphaGo and South Korean Go master Lee Sedol.

The battle between machines and humans in intellectual games has been going on for decades. Is there any special significance this time?

Artificial intelligence first shocked the world 19 years ago when IBM's Deep Blue computer defeated Russian chess king and world chess champion Kasparov. That was probably the first time in human history that another form of intelligence challenged us. After excitement, surprise, frustration, and fear, the last bit of pride was placed on Go. Some people say that this is the last high ground of human wisdom.

Finally, the "enemy" began to invade this "last stronghold of human wisdom". On January 28, the artificial intelligence AlphaGo defeated the European champion professional Go player Fan Hui 2nd Dan with a score of 5:0 in a completely fair situation. So, there is the following story.

At this time, Lee Sedol is no longer a player of any nationality or the captain of any team. He represents one side and all of humanity. People from all walks of life have their own expectations and predictions about the outcome of the game. Judging from the current situation, what are the chances of Lee Sedol holding his ground?

Known level: Lee Sedol is stronger

Lee Sedol was born in 1983. He became a professional Go player in 1995, a second-dan player in 1998, a third-dan player in 1999, and a sixth-dan player in 2003 after winning the LG Cup. In April 2003, he won the KT Cup, the largest chess tournament in Korea, and was promoted to the seventh-dan player. In July 2003, he won the 16th Fujitsu Cup and was promoted to the ninth-dan player. He has more than ten world championship titles. No one would object to his being the representative of the world's top level in the Go world today.

AlphaGo, a Go artificial intelligence program developed by Google's DeepMind, has a record of 494 wins in 495 games against similar programs Crazy Stone and Zen. When playing with a handicap of four stones (letting the opponent play four stones first), it can still maintain a 77% and 86% victory rate against Crazy Stone and Zen. It can be said that AlphaGo is the most powerful player in the field of Go artificial intelligence at this stage.

Since they are all the best of their respective "species", the question is, measured by human standards, what level is AlphaGo at?

Fan Hui, who was defeated by AlphaGo, is a Chinese professional second-dan player, and there is a huge gap in chess skills between him and Lee Sedol. However, judging from the match between the two, AlphaGo has already possessed chess skills that are no less than those of a professional first-dan player, which is also the view of most Go professionals. Ke Jie, a rising star in the Go world, is more critical in his evaluation, "Its level is probably the level before the professional stage. Although it is not yet professional, it is infinitely close to professional." Lee Sedol himself believes that AlphaGo's chess skills are equivalent to those of a third-dan player. This is already a very high evaluation. After watching the game between Fan Hui and AlphaGo, he believes that he can definitely beat the computer.

Judging from previous battle results, Lee Sedol's strength is obviously superior to AlphaGo.

Room for improvement: AlphaGo has more room for improvement

This month's match was nearly five months after the match between AlphaGo and Fan Hui. So, what changes have taken place in the levels of the two during this period of time?

In an interview with Korean media, Lee Sedol said that he would watch ALPHA's games and make some targeted preparations. However, as a mature professional chess player, it is difficult for Lee Sedol to improve his level in a short period of time.

At the same time, AlphaGo is constantly stepping up its own learning. In January this year, the Deep Mind team revealed that AlphaGo completed 1 million self-games in four weeks, which means that it can play 30,000 games with itself a day. Kim Jin-ho, director of the Big Data MBA at Seoul Institute of Science and Technology, emphasized that in order to defeat the European Go champion Fan Hui, the second dan, in October last year, AlphaGo was fed with 160,000 game records of "European amateur masters" instead of professional players' game records. Since October last year, Alpha GO has been learning based on the highest level of professional game records. This means that when AlphaGo goes to battle in March, its actual level will impress us.

The biggest breakthrough in artificial intelligence is that scientists have given machines the ability to self-learn. Perhaps this will also become its biggest potential threat to humans.

On-the-spot performance: Each has its own advantages

The existence of uncertainties increases the excitement of the competition, which is why the game is so interesting. So, in terms of on-the-spot performance, who has the advantage between AlphaGo and Lee Sedol?

A big advantage of machines is that they have no emotions. They don't have to bear the pressure of a man-machine battle, nor do they have the mentality of having to win. Fan Hui once mentioned that in the game with AlphaGo, part of the reason for the defeat was that the psychological burden was too heavy and they were too anxious. Judging from past records, AlphaGo's performance is very balanced, and it is almost impossible to see the machine's stiffness, and there are few "failures" that other Go programs have. Although the game on March 9 was set in Seoul, South Korea, it is difficult for Lee Sedol to say that he has any home advantage. On the contrary, if AlphaGo's progress is really great, even far beyond Lee Sedol's expectations, it is hard to guarantee that it will cause a psychological impact on him.

However, Lee Sedol's style of play is known as the "zombie style". In the middle of the game, he often makes excessive moves, resulting in the death of the dragon. In the eyes of professional masters, such a one-sided situation is the end of the game, especially for the old-school Chinese and Japanese professional chess players who only study the first half of the game. However, Lee Sedol can often burst out amazing power in desperate situations. He returns to full health again and again like a resurrected zombie, and always crosses the finish line at the last moment. Playing cards that do not follow common sense may be the most difficult part for AlphaGo to grasp Lee Sedol.

Although AlphaGo relies on a combination of precise expert evaluation systems, deep neural networks based on massive data, and the traditional artificial intelligence method Monte Carlo tree search, and has the ability to self-learn, it is ultimately unable to escape the limitations of previous human experience and it is difficult to create a new style of play.

It can be said that in terms of performance, AlphaGo and Lee Sedol are evenly matched.

In terms of the expected results of the game, Chinese netizens are more inclined to vote for South Korean player Lee Sedol. Obviously, this game has long broken the restrictions of national boundaries. For artificial intelligence scientists, they have always maintained a mentality of both curiosity and fear. The significance of this game to humans is far more than just winning or losing.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.