AlphaGo has already won over the European champion in Go, allowing another victory does not seem impossible. Claes Strannegård, associate professor of cognitive science at Chalmers says that it is probably only a matter of time before a machine is superior to people of Go.
– It becomes a very important milestone in AI history. It has been sooner cracked the Lyg yarn important games, chess, damspelet, and quite recently even poker. What has withstood far is just the game Go, says Claes Strannegård.
The reason that Go has been so difficult cracked AI computers is that the game contains so many possible positions, then the board in its largest model consists of 19 x 19 lines. That now seems to approach a computer to crack a human in Go, because according to Claes Strannegård partly because it has developed the machine’s ability to think abstractly, which has made it approaches people’s thinking.
– It is two techniques combining to achieve this result; Deep learning and reinforcement learning. It is not based on counting out lots of variations of the standings, but more that you look at a situation – that I move the stone here compared to there – and evaluate how good it is. The machine has a better opinion without having to watch so many steps forward, says Claes Strannegård.
How have you been there?
– Yes, it here computer Go has played millions of times just so that when the score looks a certain way, you know AI if it is positive or negative. It’s a very good algorithm in the bottom, but without experience becomes worth nothing.
What can this new technology is to be for the benefit off the board game world?
– It may for example include robots that can learn how to get around in difficult terrain – which Google is also doing – so for example, they can make the rescue. Then we have in itself boring uses as well as robotic soldiers.
Like New Technology on Facebook
No comments:
Post a Comment