Hacker News Reader

Comment by badminton1 | Hacker News Reader

Some professionals labeled some AlphaGo moves as being unoptimal or slow. In reality, Alpha Go doesn't try to maximize its score, only its probability of winning.

View on HN

Right, but I think the initial concerns have been proven wrong in some respect. I think AlphaGo can accurately determine when a move in a local area doesn't significantly change the outcome and then just does something else for a different reason. AlphaGo has a higher willingness to go somewhere else, and I think that's the algorithm finding another path that is more helpful than the current one for 1 move.

Humans have a tendancy to want to win the battle, or to get too focus in a local area. I think that's a way AlphaGo is coming up with an extra move here or there which is making a difference in the fight later.

From watching it I'm almost inclined to say it maximizes its chances of not losing over necessarily winning.

They are playing with a 7.5 point komi [1], and so a game cannot end in a tie. Doesn't that mean that there is no distinction between winning and not losing?

[1] komi is a number of points that is added to white's score to compensate for the disadvantage of moving second. When a non-integer komi is used, such as in this match, you cannot have a tie because scoring on the board is always integral.

The developers have said that. They also said it doesn't care if it wins by 20 or by 1. It just maximizes winning at all.

Your comment reminds me of the Star Trek TNG episode Peak Performance. Data wins the rematch by playing to stalemate.

https://en.wikipedia.org/wiki/Peak_Performance_(Star_Trek:_T...

Sorry, but what's the difference?

There really isn't, from the point of view of AlphaGo. But for someone playing Go (or any game), it's a hint that perhaps one should focus on things that can go wrong more than furthering your own aggressive plan.

It's certainly a feature of the best Magic: the Gathering pros, for example - their play is marked by the cards they play around, even when seemingly far ahead.

interesting that "not losing" (avoiding mistakes that lead to "blowing up") also seems to be common philosophy shared among very successful investors of wildly differing styles, who are playing another sort of game in the markets

Yeah, isn't Lee Sedol playing with the same objective? It sounds like people are implying he's trying to thrash the AI. Or do they mean he can't help but have such an objective, subconsciously?

The issue is more that humans are not quite so confident of their positions as AlphaGo sometimes can be. If you are 100% sure that you're ahead by 1.5 points, you would play very differently than if you're ahead by 1.5 plus or minus five.

Through watching those few games, parent has deduced that AlphaGo prefers ties

With komi at 6.5 or 7.5 there are no ties.

Reply on news.ycombinator.com