The story of AlphaGo so far

During the game, AlphaGo has played a very creative victory, some of them - including 37 points of the second game - -, surprisingly they overthrew hundreds of years of wisdom It was greatly reviewed by players of all levels. In the process of victory, AlphaGo taught a whole new kind of knowledge to the world. It is probably the most studied and most anticipated game in history.

Since then, AlphaGo was amazingly amazing. In January 2017, the improved AlphaGo version was released as the "master" of online players and won the top 60 international game high-speed time control games with the top international Go player for 60 consecutive victories.

In May 2017, Alpha Go participated in "Future Journey" of the birthplace of China Go and studied Go mystery deeply in the spirit of mutual cooperation with domestic top players. For details of Five Days Summit, please click here.

Five months later, we received another Nature paper from AlphaGo Zero. Unlike previous versions of AlphaGo, which learned how to play games using thousands of amateur and professional games, AlphaGo Zero learned to start a completely random game and play Go games .

By doing so, it surpassed all previous versions, including people who broke Lee Cedl and Ke Gee of World Go Champion.

We believe that this new breakthrough can promote great scientific breakthroughs and doing so will bring a revolution to the world.

Learn more about AlphaGo and do it behind the scenes of Seoul using AlphaGo movies. Let's now see Netflix, Google Play Movies, Amazon Instant Video, iTunes

AlphaGo Zero is the latest version of its Go-playing automation. Some people think that it is difficult to exceed the AlphaGo version that won the Human World Championship with Go. However, AlphaGo Zero broke not only the previous system but also the way to verify innovative ways. More specifically, this is what AlphaGo can accomplish. The first point of many people does not appear to be important. Probably because the gradual improvement of technology is always the standard. Perhaps there is a possibility that one algorithm directly invokes another algorithm within another algorithm 100, and another person does not have the same appeal that it is 100 times the same as humans. There is no contradiction in this algorithm as seen by humans.

Humans learn languages ​​through metaphor and stories. The human strategy found in Go is referenced by name so that the player can recognize it. Go 's human language can be inefficient as it can not express more complicated complex concepts. What AlphaGo Zero can do is to move it to meet multiple goals simultaneously. As a result, early versions of human beings and AlphaGo were limited to relatively linear thinking, but AlphaGo Zero was not plagued by an inefficient policy language. Interestingly, this may seem to be a system that does not actually use the implicit bias that may exist in the language. DeepMind's David Silver has a more bold proposition.