Developers of AI "Deep Blue" who defeated chess world champion about "AlphaGo" who defeated go game champion talks

ByBuster Benson

AlphaGo and Lee Se-dul (Lee Seung-kyu) Kuroda's opponentAs was the case with artificial intelligence (AI) and many world top players of intelligent games like Go, chess, and chess, so far had many conflicts, it was the chess world champion at the timeGarrie KasparovIBM's "to say even as a founder of the AI ​​top player who got a lot of attention from all over the world by killing Mr."deep blue"Answers to interviews about AI and AlphaGo.

Deep Blue developer speaks on how to beat Go and crack chess | The Verge
http://www.theverge.com/2016/3/12/11211306/ibm-deep-blue-murray-campbell-alphago-deepmind-interview

Artificial intelligence (AI) development startup acquired by Google over 50 billion yen in early 2014Deep Mindis. Google has developed DeepMind's AI technology, and in order to challenge a professional shogi player of Go which could not be beaten by computer software so far, AI software "AlphaGoDeveloped. And in March 2016 AlphaGo won the top game of Lee Se-dul Nine stage in the world of Go. The game is the fifth game against the other, and AlphaGo's already won 3 wins and 1 loss at the time of writing the article, the losing of the Cedor Kudan is confirmed.

Just 20 years before this historical achievement, IBM developed AI's Deep Blue won the chess world champion Garry Kasparov. Mr. Campbell, Mr. Campbell, who is one of the members who developed that deep blue, is still working at IBM and is a senior manager of the cognitive computing division, which develops IBM's AI "Watson". Mr. Campbell talks about Deep Blue and AlphaGo of Google.

ByBlume Fou

Mr. Campbell said about deep blue development process at the time, "We thought that pure powers could not beat the world champion of chess and instead adopted the approach of using enormous computing power to create a difference "" So, I introduced AI type advanced algorithms and super computer level machine power, and succeeded in creating AI software of chess world champion level. "

Also, in response to the question "How much chess knowledge was needed" in Deep Blue development, while preposing that "knowledge of chess to a certain extent is necessary", while the development side is a strong chess player It did not have to be, and in reality developers reveal that no one was an excellent chess player. However, as development of Deep Blue came to the final stage, generalGrand MasterIt seems that it was necessary to investigate details of what kind of exercise or how to practice chess player (excluding world champion, highest title of chess player), in that case, professional athlete grandmaster Joel Benjamin It seems he got advice from players and others. In addition, it was necessary to have the chess experts play against each other in order to ascertain the completion of Deep Blue.

Deep blue now has been disassembled, but some of the used racks are exhibited in the computer history museum.

ByJim Gardner

Many chess players were surprised at AI's extraordinary hands when Deep Blue defeated Mr. Kasparov in 1997, and hands that are unlikely to be hit by such humans are called "computer move" It is said that. Campbell says that "computer move" is also considered to be a major factor that AI won the chess professional player, "I can not read the flow of the game because of a very abnormal hand." . However, the next generation of young grandmasters are more capable and capable of handling "computer move" than grandmasters who have taken ages. This is a proof that the chess player grew by playing against the computer, Mr. Campbell said.

As to the question "From the computer science point of view, what is the difference between go and chess?", I have never hit Go but I pre-emphasized that I have a lot of knowledge While commenting, "It is difficult to measure the situation of players over chess." Whereas Go plays a game through dialogue with opponents, Chess has a different nature that a piece moves on the board, so if you see which player has which piece, It is easy to predict whether it is dominant. However, it is not necessarily that the game is proceeding to the dominant position because the Go has not necessarily many stones, so it seems that evaluation is difficult.

ByKenming Wang

And the question "What do you think about AlphaGo?" Says, "AlphaGo is a proof that it is an advancement of state-of-the-art technology that clearly shows that not only the general mechanism of Go but also that of every game I am learning, it is very moving. " In addition, if you create an AI for chess with the same approach as AlphaGo, I will also say that AI will be created that can defeat all Grand Masters.

However, it is said that it can not be a state-of-the-art technology, because the state-of-the-art chess program is incredibly strong and far superior to humans. And why such an incredibly strong program is born in the chess world is that Chess is an important game in which "reading" is more important than Go, Mr. Campbell said. It seems that Mr. Campbell thinks that it is important to measure how the opponent moves from intuition and estimation rather than deep reading.

In addition, Mr. Campbell responds to questions such as "Which of AlphaGo and Cedor Kuroki wins?" And "How important is machine power?" In the interview.