AlphaGo beats Lee Sedol again in match 2 of 5 • Hacker News Board

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#241

Post by conanbatt » Thu, Mar 10, 2016, 1:34 PM UTC

What is interesting to me is that the computer makes clear mistakes when its on the lead. Since it might find the chances to win equally among different scoring results, it often picks a weaker one.

This has a powerful consequence: we have not seen AlphaGo pushed to the limit, he is lowering the distances as if it were playing a teaching game.

Lee Sedol I think came to this conclusion, and the only human strategy left is to take a lead big enough to maintain the rest of the game. And that might be the last strategy to play to show the computer is already unbeatable, because it will be pushed to its limits to win a game and it might overcome humans.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#242

Post by dfan » Thu, Mar 10, 2016, 1:35 PM UTC

drjesusphd wrote: post #180

So is this it then as far as games go? Does anyone know of any efforts to develop a more "human-friendly" complete information game than go?

The game of Arimaa was created in 2002 specifically to be a complete-information game that computers couldn't beat humans at.

As of 2015, the best player in the world is a computer.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#243

Post by taneq » Thu, Mar 10, 2016, 1:35 PM UTC

mathgenius wrote: post #64

IvyMike wrote: post #25

I sense a change in the announcer's attitude towards AlphaGo. Yesterday there were a few strange moves from AlphaGo that were called mistakes; today, similar moves were called "interesting".

AlphaGo maximises the probability of winning, and not the margin by which it does. So those "mistakes" yesterday turned out to be fortifying moves because AlphaGo was confident of a win. And similarly today the weird moves were interesting because they perhaps indicated that AlphaGo thought it was ahead.

Yeah, that explanation from the DeepMind team member today put a whole new spin on some of the 'odd' late game moves. It doesn't 'care' about about margins so it will shore up its odds of a win in preference to increasing the margin if it wins.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#244

Post by lmm » Thu, Mar 10, 2016, 1:37 PM UTC

crispyambulance wrote: post #215

lmm wrote: post #125

Earlier quoted context omitted.

How soon we forget. Twenty or thirty years ago Chess was spoken of in exactly those reverent tones.

I suppose in the era of Bobby Fischer, chess was proxy for superpower one-upmanship. That's long gone, but Chess as a game is still doing fine and I expect that it will be the same for Go. We still have chess tournaments, super-star grandmasters and circus freaks (people who can play blindfolded against multiple opponents). And, yes, computers can easily smoke all but elite players. Why should Go be different?

I played Chess as a kid. I watched the local tournament shrink from the big town hall, to the side room in the same building, to a local school hall, to a classroom in that school. I really do think the game is dying - perhaps something that was happening already, but Kasparov losing to Deep Blue seemed to really catalyse it.

(Not saying this is a bad thing. Evolution in games is natural, and I think it's amazing how much innovation is going on right now (particularly enabled by Kickstarter) - you'd think that board game design would have been worked out decades or centuries ago, but in the same way that incandescent bulb development accelerated massively when competition arrived, it feels like game design has got so much better when forced to compete with computer games. If there are other activities that people find more fun than Chess, that's all to the good)

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#245

Post by astrofinch » Thu, Mar 10, 2016, 1:37 PM UTC

So given that this victory seems to be happening a decade or so before experts predicted, how likely are we to see similar acceleration in reaching other AI milestones? (Especially given that AlphaGo is using the same algorithm that won the Atari games, so it has the potential to be very general in its application)

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#246

Post by boucher » Thu, Mar 10, 2016, 1:38 PM UTC

EliRivers wrote: post #173

Smaug123 wrote: post #153

Earlier quoted context omitted.

If a parent watches their child win a game, would that be them achieving the win themselves? If so, why does having nine months of creation time and ten years of training time change that?

There is no child here. There is a deterministic set of calculations. The developers created a deterministic set of calculations, and then executed those calculations. AlphaGo is not a child. AlphaGo is a set of deterministic calculations, defined and created by some people. Your question about humans and their children is irrelevant. There is no child here. In this flawed analogy, the "parents" are the developers, a…

And yet the operative fact is that a human could not execute those calculations in a lifetime.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#247

Post by drjesusphd » Thu, Mar 10, 2016, 1:38 PM UTC

dfan wrote: post #242

drjesusphd wrote: post #180

So is this it then as far as games go? Does anyone know of any efforts to develop a more "human-friendly" complete information game than go?

The game of Arimaa was created in 2002 specifically to be a complete-information game that computers couldn't beat humans at. As of 2015, the best player in the world is a computer.

Whoa. Well, I guess that's a losing battle.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#248

Post by taneq » Thu, Mar 10, 2016, 1:39 PM UTC

conanbatt wrote: post #241

What is interesting to me is that the computer makes clear mistakes when its on the lead. Since it might find the chances to win equally among different scoring results, it often picks a weaker one. This has a powerful consequence: we have not seen AlphaGo pushed to the limit, he is lowering the distances as if it were playing a teaching game. Lee Sedol I think came to this conclusion, and the only human strategy lef…

I don't see how you can term them as "clear mistakes" when the game is playing at a higher level than any of us meatbags. In this case (according to DeepMind) it's no different to a racing driver backing off the pace in the last few laps if they have a big lead - it's better to guarantee a win than to win by a large margin.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#249

Post by Jach » Thu, Mar 10, 2016, 1:40 PM UTC

taneq wrote: post #232

eru wrote: post #135

Earlier quoted context omitted.

Fan Hui has already learned from AlphaGo. He's been playing matches against her regularly, and (perhaps as a result of that) won _all_ his games in the last European championship.

I find it interesting that people are using gendered pronouns for AlphaGo. Getting some definite Turing-test vibes here.

Myungwan Kim said he feels like it plays like a she. Personally I think its informal gender was determined by the nigiri of the first match, as it's common to refer to black as he and white as she absent of player names. And I'm expecting to see at least one really cute AlphaGo-tan drawing any day now.

Re: AlphaGo beats Lee Sedol again in match 2 of 5

#250

Post by pavelrub » Thu, Mar 10, 2016, 1:41 PM UTC

liviu- wrote: post #195

pavelrub wrote: post #134

Earlier quoted context omitted.

This doesn't seem to be the main reason why Go is harder than chess for computers. It was noted that even in 9x9 Go, with a comparable branching factor to Chess, traditional Go programs are still no stronger than on big boards. The main difficulty for Go is that it is much harder to evaluate board positions. So in Chess the depth of the search can be significantly reduced by using a reasonable evaluation function, wh…

>It was noted that even in 9x9 Go, with a comparable branching factor to Chess, traditional Go programs are still no stronger than on big boards. Are they not? MoGo beat pros of 9 Dan on 9x9 in 2011: https://www.lri.fr/~teytaud/mogo.html

Well, I guess it was more true before the advent of Monte Carlo Tree Search. Even so, note that even in the case of MoGoTW in 2011, it played blind Go (this helps the computer), and out of 4 games, won two games against a 9p player, and lost 1 game to a 5p player. Though it is perhaps better than MoGo's performance on 19x19, it still isn't very good, doesn't seem much better than MoGo on 13x13, and performs much worse than computer Chess, despite a similar branching factor.