Quote:
Originally Posted by abysmal01
I found it kind of hard to believe that in thousands of years of playing the game no one had realized that sometimes you'll increase your chances of winning by playing to win by 1 point instead of 20 (which I think is what they were saying ag taught them).
Unless I'm remembering incorrectly this is not what they were saying. What they were saying was AlphaGo only cared about making the move that would statistically give it the best chance to win the game.Some of these moves looked odd and would sometimes cause AlphaGo to lose in weird amateurish looking ways but they were still the best moves.
If AG was in a position where it was going to lose the majority of the time it would aggressively try and change these odds, not sit back passively and hope the opponent made a mistake it could capitalize on. AGs ability to calculate these odds and run simulations to determine the best move is why it could make moves, such as the one we see in the doc, that a human just doesn't have the computing power to make.
"Drawing on its extensive training with millions upon millions of human moves, the machine actually calculates the probability that a human will make a particular play in the midst of a game. "That's how it guides the moves it considers," Silver says. For Move 37, the probability was one in ten thousand. In other words, AlphaGo knew this was not a move that a professional Go player would make.
But, drawing on all its other training with millions of moves generated by games with itself, it came to view Move 37 in a different way. It came to realize that, although no professional would play it, the move would likely prove quite successful. "It discovered this for itself," Silver says, "through its own process of introspection and analysis."
Is introspection the right word? You can be the judge. But Fan Hui was right. The move was inhuman. But it was also beautiful."