Two things that Google learned from the game of artificial intelligence AlphaGo and the world's strongest shogi player
The artificial intelligence development venture acquired by Google "Deep Mind"Artificial intelligence software developed by"AlphaGo"Is the world's strongest shogi player with a go that artificial intelligence has never wonLee Se-dul (Lee Seung-kyu) break through KudanI accomplished a fruitful achievement. In a game in the century of the Go held in Korea, although the result that artificial intelligence exceeds humans has come out, Mr. Demis Hasabis led by DeepMind said that victory of artificial intelligence has great profit for all human beings We emphasize.
Official Google Blog: What we learned in Seoul with AlphaGo
AlphaGo and Lee · Cedor Kuroda 5 strikeouts, which took place from 9th March to 15th March 2016,AlphaGo's 4 wins and 1 lossThe settlement arrived in the form of. The battle of the century was delivered worldwide by live broadcasting, and Google explodedly increased the number of go-related searches such as Go rules and go boards,Popularity of Kore in KoreaIt has attracted a lot of attention, such as the fire burning and the hashtag "Man vs. Machine Go Showdown" breaking through 200 million page views in Weibo of Chinese social media.
The interest of many people was suitable for the composition of "artificial intelligence VS human beings, which is win", but for Mr. Hasabis who is also a developer of AlphaGo it is not only the victory of artificial intelligence was sought. Prior to being acquired by Google, Mr. Hasabis founded DeepMind to learn himself and eventually to develop artificial intelligence as a tool to solve problems surrounding human beings and society, such as environmental problems and disease diagnosis .
AlphaGo uses a neural network that imitates human neural circuits to perform data classification "Deep learning"Try an action and learn better behavior from the profit obtained from the result"Reinforcement learningIt is software developed using. The first big AlphaGo challenge made public in January 2016 was a battle confrontation with Lee Se - dol Kuzu, the strongest player of Go, which is said to be the most complicated in board games.
AlphaGo has struck a hand with a commentator of Go to 'Lee Cedor Kuroda', which praises 'unpredictable', 'creative', 'beautiful'. According to DeepMind's game data, AlphaGo's in the second station37 th handIs one that was analyzed as 0.001% possibility of human beings hit. Meanwhile, Lee Se-dul Kuroda released at the fourth station78 th handIt is analyzed that the possibility that a human being hits is 0.001%. Lee Se-dul Kuroda commented after the game at the 1st station that "AlphaGo's strategy in the early stage is outstanding and AlphaGo hit a non-ordinary hand that people will not hit."
Mr. Hasabis says there are two things I got at the AlphaGo and Lee Se-dul Kudan game. One was able to test the ability of artificial intelligence to solve various problems. AlphaGo expanded his vision to the whole board, hit a hand that was taught that a professional chess player should not hit, or did not think to strike. This is because there are many possibilities that artificial intelligence can be used in other fields other than Go, because artificial intelligence has creative problem solving skills unexpected by human ability.
The second is that the game is told by the human being that the game is "human VS machine", but AlphaGo is made by humans, AlphaGo's victory is a human achievement. AlphaGo and Lee Se-dul Kuhno release one of the new ideas and one to solve the problem in the game, and this is thought to bring great benefits to human beings when viewed from a long-term perspective.
However, Mr. Hasabis also emphasizes that this victory is only a small step for artificial intelligence, as it is said that in Korea 'Go' or 'losing' should not be ignored in Go 'in Go. It seems that it will take a long time for artificial intelligence to be able to do tasks that human beings use with their heads.
in Software, Posted by darkhorse_log