Discussion of AlphaGo's Recent Streak

I am surprised that no one has started a thread to discuss the recent AlphaGo (AG) streak. Reactions, observations, questions? Okay, I’ll get the ball rolling. I have played through 21 of the 60 games (is this still the current number?): the four at AGA, and #1-16, plus #40 (Sedol), on the German site. I fully understand my severe limitations as a lowly DDK, but I will venture some observations nevertheless, at least in the hope of provoking comment (“Fools rush in, where angels fear to tread”). Tell me if I’m wrong, what I’m missing (lots, no doubt), or whether I’m merely quotidian.

AG is relentlessly, frighteningly aggressive in its play. It seems to play for territory with light, distributed moves rather than for influence by thickness (isn’t this an older style of play?). Most important, AG systematically and rapidly blocks every access to the center, and the wonder is how successful it is at doing this. Often, the human player is largely shut out of the center by 50 moves! (Sedol and a couple others did a good job of avoiding this.) These games were played in a short span of time, and I would like to hope that the pros would have won some games had they had a chance to study AG’s style (since this is a new iteration of the program, I discount AG’s earlier games as a guide to its present style). Future games will show whether this hope is realistic or not.

I finish with a specific practical question. In game 9 (vs. Meng Tailing as white), on move 107, AG could have saved its group of 7 stones and simultaneously captured 4 stones by playing F7, but it played elsewhere and lost the group. AG’s actual move didn’t seem worth the loss, so either I’m not understanding something, or AG made an oversight (is that even possible?).

Have at it, folks!

3 Likes

All I have to say is if I played like this (see below, Alphago is W) and had it reviewed by a 9 dan, I would be spanked mercilessly. Yet, the new Alphago is undefeated.

“You know nothing John Snow.”

2 Likes

From my perspective the answer would be Master§ put more importance in breaking into whites center over saving those stones. Which I agree point wise it was worth more, To break into the center like Master§ did.

Imagine If black diid capture those stones they would get 8 points from the capture and with it beings whites turn they can cut off the center.

This Game: https://online-go.com/game/7468001

Thank you. Yes, I see that now. Ironic that AG had to worry about that, when usually the shoe is on the other foot.

Keep in mind that when it plays questionnable moves, it can simply be because it doesn’t see it reducing the probability of winning, like if it think it’s already won.

1 Like

@xavierm02 has an excellent point, keep in mind that AlphaGo is selecting the safest path to victory, not the one that maximizes the score.

1 Like