According to @CCTV CCTV5, Google's Deepmind company officially announced at the DLD Innovation Conference in Munich, Germany, that it will launch the real 2.0 version of AlphaGo.
The new version of "Alpha Dog" will officially abandon the human game, relying on the computer's own deep learning to grow - that is, to explore the limits of the "deep learning" approach, to the realm of "God of Go".
The 2.0 version of AlphaGo will feature Go-Arts and Deep Learning, which are truly artificial intelligence (AI).
Google executives meet Nie Weiping
Ke Jie Weibo
In the beginning of 2017, the Alpha Dogs of the Incarnation Master completed 60 wins, and the only draw was because the players were dropped.
At that time, "Chess St." Nie Weiping said: "Go is far from being as simple as we think. There is still a huge space waiting for us to dig. The Alpha dog, Master, or the "Go God" sent to Humans lead the way."
Ke Jie said that he had no sleep all night. "The human race of the millennium has evolved, but the computer tells us that human beings are all wrong..."
Since the wisdom of human Go is now in the doldrums, what kind of Go change will artificial intelligence bring?
DeepMind has long had the idea of ​​abandoning the human game and taking the road of AI.
Father of the Alpha Dog: Hasabis
The company's founder, Hasabis, previously revealed in an interview that their team is trying to train an artificial intelligence that has not learned human chess.
When the Master killed the Quartet, there was also speculation that the Master had never studied the human game.
But whether it is a human routine or a self-contained faction, the Alpha dog is even more terrible is its ability to learn.
In October 2015, the Alpha dog won 5-0 in a fair situation with the European champion Fan Wei. In an interview with the media, Fan Wei believes that the fear of Alpha dogs lies in "never make mistakes."
Nie Weiping and Master play
After Fan Yi and Alpha Dog's five-set battle game was put on the network, there were quite a few amateurs in the game who commented on the Internet. Many of them thought that Fan Hao's game was not good.
But at the time, Nature wrote that "humans have a disadvantage when playing chess. After a long game, they make mistakes, but the machines don't. And humans can play 1000 games a year, but the machine can play in one day. 1 million innings."
"The learning ability of Go AI is very strong, just like the original 'Deep Blue' lost to Kasparov for the first time, but won the game in the second game less than a year."
The Korean chess house awarded the honorary nine-segment certificate to the Alpha Dog.
According to Nature, the core of Alpha Dogs is two different deep neural networks: the "policy network" and the "value network". Their mission is to collaborate to "pick out" those promising moves and discard obvious bad moves.
What is even more horrifying is its "evolutionary" deep learning ability, which uses 20 million innings to train the game, and each time it can improve the game.
72V 125Ah Power Battery,Portable Solar Power Station,72V Camping Power Battery,Rv Mobile Energy Storage Power Supply
Sichuan Liwang New Energy Technology Co. , https://www.liwangnewenergy.com