AlphaGo Zero Learns Without Human Involvement
AlphaGo beat the best Go player in the world, the entire Internet Go community, and the Korean champion. But the game has been learned by analyzing thousands of played games among the best human players.
AlphaGo Zero only got the rules of the game, then he learned the game himself, so he played against himself.
The results are astounding. As AlphaGo Zero advanced from the perfect beginner to the capacity of the version that last year overtook Lee Sedol, in 21 days he became better than the version that this year beat the world champion, and in 40 days he became the best artificial intelligence to play the game.
AlphaGo Zero achieved all this by itself, without seeing how people play. This proved to be an advantage since it was not burdened with fixed patterns but developed some new ones that we have not seen in the past.
In other words, this means that artificial intelligence has learned more in a month ago than humanity in millennia. A science article about that is here.
A video where DeepMind's Professor David Silver describes AlphaGo Zero:
More about that in DeepMind's Blog.
@steem-marketing has voted on behalf of @minnowpond.
If you would like to recieve upvotes from minnowponds team on all your posts, simply FOLLOW @minnowpond.