Life In 19x19
http://www.lifein19x19.com/

Predict AlphaGo on Future of Go Summit in Wuzhen, China
http://www.lifein19x19.com/viewtopic.php?f=18&t=14159
Page 3 of 7

Author:  pookpooi [ Fri May 19, 2017 4:51 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Official website launched
http://events.google.com/alphago2017/

Author:  pookpooi [ Sat May 20, 2017 7:46 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Rumor around weibo that not only the new AlphaGo will learning without human game record, but it also doesn't use Monte Carlo Tree Search

http://www.weibo.com/p/23041873040b820102wwci

My opinion, is this even possible? Please take this as a grain of salt. But there's information that DeepMind dedicated the whole day (24th) to conference. (Is someone roaming through DeepMind presentation slides without permission? That's very bad)

Author:  alphaville [ Sat May 20, 2017 10:05 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

pookpooi wrote:
Rumor around weibo that not only the new AlphaGo will learning without human game record, but it also doesn't use Monte Carlo Tree Search

http://www.weibo.com/p/23041873040b820102wwci

My opinion, is this even possible? Please take this as a grain of salt. But there's information that DeepMind dedicated the whole day (24th) to conference. (Is someone roaming through DeepMind presentation slides without permission? That's very bad)


Any chance you can post the weibo info without the need for us to login there?

The Nature paper stated that neural-network only (no MCTS) version of AlphaGo was already stronger than all other commercial computer-go games at the time of Fan Hui match, but way weaker than the MCTS version.
It sounds impossible at first sight to think that a pure static analysis version, without MCTS, would be stronger than a pro; on the other hand, with a very deep neural-network architecture, I guess it may be possible, since some of the layers of the network would effectively do some "reading ahead". Just speculating here.

Author:  pookpooi [ Sat May 20, 2017 11:55 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

For me it doesn't require sign up to access weibo, it just that they're in Chinese language, but here in spoiler.

柯洁与AlphaGo的比赛又要开始了,这次有什么看点呢?想了几天还没有想出来。并不是没有看点,而是谷歌一直没有透漏,或者说我没有看到,今天的AlphaGo与去年的AlphaGo有什么不同,不知道有什么不同,就不知道看什么。

以前在文章中我曾经猜想过,也许新的AlphaGo会尝试从零学习,也即抛弃从人类棋谱中学习,完全靠自我互博,逐步提升自己的能力,但是一直得不到证实或者正否。我一直觉得这是有可能的,不过可能需要的计算资源要大的多,虽然一直有朋友反对我这种看法。

前几天和另外的朋友聊天时,得到了更加激进的消息,说新的AlphaGo不仅是从零学习,而且还舍弃了蒙特卡洛树搜索,让我大吃一惊,如果真是这样,那么将是又一次大的进步。不过我总有点不太相信,从零开始是有可能的,但是舍弃蒙特卡洛树搜索,还是有点不可想象。

前几天看到聂卫平的文章,提到有些棋手已经在跟AlphaGo学习,在比赛中走出了之前认为不好但是AlphaGo走过的类似的棋,我想这就是AlphaGo对围棋发展所起的推进作用,如果AlphaGo真的舍弃人类棋谱,从零开始学起的话,这种贡献可能会越来越多。

期待在这次比赛之后谷歌团队能透漏一些技术上的问题,这样对于我们如何看待这次比赛,会有很大的帮助。

总之这次比赛关注的已经不是谁输谁赢问题了,而是技术上是如何发展的,这对AI才是有意义的,否则如果只是重复过去,则意义要大打折扣。

Author:  Uberdude [ Sat May 20, 2017 1:25 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

It's important to note that saying this new version of AlphaGo doesn't do Monte Carlo Tree Search is not the same as saying it doesn't do tree search, though I don't know if that technical precision was lost in translation. You could have a tree search version which constructs a game tree using a policy network to suggest moves, and a value network to evaluate nodes in that tree without doing monte-carlo rollouts to the end of the game. In their published work so far combining the rollouts with the value network gives the best results, but as the networks keep getting better it wouldn't surprise me if tree+policy+value+rollouts was now 15p strength and tree+policy+value "only" 13p and therefore good enough for the upcoming match.

Author:  djhbrown [ Sat May 20, 2017 3:49 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

anyone got a spare hat to eat or talk through?

hats off to the PorkyPie guys who know a good Bernays story to get the chattering classes chattering

no tree search... ho ho ho... sometimes people can't see the forest for the spin-fluff. [correction: delete 'sometimes' insert 'always' eg Persil washes whiter, Coca-Cola refreshes you, Bay of Pigs, Bay of Tonkin, Baywatch, government by the people, in God we trust, etc].

check out the esteemed opinions of L19 experts this time last year - about as reliable as a pre-Brexit poll.

so far Swim is the only bot to have bettered the dog, and with < 1 googleplexth of the dosh too

Author:  alphaville [ Sat May 20, 2017 3:51 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Uberdude wrote:
It's important to note that saying this new version of AlphaGo doesn't do Monte Carlo Tree Search is not the same as saying it doesn't do tree search


The way I read it, it was just that, no real read-ahead but just:
1. Policy net to generate some good candidate moves.
2. Value net to select the one leading to most favorable position after the next-move.

Author:  Uberdude [ Sun May 21, 2017 2:02 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

I know I shouldn't bite, but here goes...
djhbrown wrote:
so far Swim is the only bot to have bettered the dog, and with < 1 googleplexth of the dosh too

Swim is not a bot: it is the beginnings of a design for a bot with a lot of implementation details outstanding. If by "bettered the dog" you mean provided a post-hoc rationalisation of the correct answer to Lee Sedol's wedge in game 4 then you are wrong: Leela plays the correct move too for example (and without a human operator who already knows the correct answer to guide it). Leela is a bot which actually exists because it has been coded and can run on its own without a human operating filling in the gaps. And is just a one-man-band too so a lot cheaper than AlphaGo.

Author:  djhbrown [ Sun May 21, 2017 4:09 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Uberdude wrote:
here goes...a post-hoc rationalisation .. Leela plays the correct move too
man bites dog, except Swim isn't a dog, she's a lion. J13...L10 was pre-hoc, not post-hoc. Leela plays L10 too? If true, that's interesting - it suggests maybe the dog out-thunk itself, and could do so again...

Oh, the excitement, the tension! Will Jie, won't he, will he, won't he, won't you join the dance? That must be worth something, porky?

Author:  Kirby [ Sun May 21, 2017 5:44 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

I missed some post somewhere, maybe, but what is the connection between "AlphaGo", "alphadog", and this entity we are referring to as "dog"?

Are these all references to AlphaGo? If so, why?

Author:  djhbrown [ Sun May 21, 2017 6:45 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

https://www.reddit.com/r/baduk/comments ... og_is_not/

Author:  Kirby [ Sun May 21, 2017 8:17 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

djhbrown wrote:
https://www.reddit.com/r/baduk/comments/4aibtg/ke_jie_said_alphago_is_a_dog_and_this_dog_is_not/


Thanks, it makes sense now :-)

Author:  Bonobo [ Mon May 22, 2017 7:46 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Totally off-topic :cool: therefore hidden.
Attachment:
The joyful, loving, eternally forgiving nature of dogs.jpg
The joyful, loving, eternally forgiving nature of dogs.jpg [ 509.57 KiB | Viewed 7883 times ]

Author:  alphaville [ Mon May 22, 2017 8:27 am ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Bonobo wrote:
Totally off-topic :cool: therefore hidden.
Attachment:
The joyful, loving, eternally forgiving nature of dogs.jpg


Ha, food for thought... :-)

Author:  Bonobo [ Mon May 22, 2017 12:39 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

(Via Vincent Lejeune’s post on FB, thank you!)



<edit>
Uhm, the blank space above is this:
Code:
[youtube]Fs1sFJ94rAU[/youtube]
Could it be that our YouTube tags still rely on http instead of on https? “Or something”?
https://www.youtube.com/watch?v=Fs1sFJ94rAU
</edit>

Author:  pookpooi [ Mon May 22, 2017 1:02 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

I actually posted that on Reddit before but decided not to post here because L19 doesn't support YouTube embed.
Or maybe it's supported but intentionally disabled by webmaster.

Author:  Kirby [ Mon May 22, 2017 1:50 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Bonobo wrote:
<edit>
Uhm, the blank space above is this:
Code:
[youtube]Fs1sFJ94rAU[/youtube]
Could it be that our YouTube tags still rely on http instead of on https? “Or something”?
https://www.youtube.com/watch?v=Fs1sFJ94rAU
</edit>


Thanks. I fixed this, so your post shows the video, now.

Author:  djhbrown [ Mon May 22, 2017 2:55 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

Never mind the quality truth, feel the spin-width

[totally on-topic so not hidden like your motives]
Attachment:
olddogsnewtricks.jpg
olddogsnewtricks.jpg [ 98.11 KiB | Viewed 7756 times ]
____________________
i stink, therefore i glam

Author:  hyperpape [ Mon May 22, 2017 6:10 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

I see that there's going to be a stream at the DeepMind site, but will there be any live commentaries? I think I'm too tired tonight, but I'll probably end up staying up to watch the beginning in spite of that.

Author:  dfan [ Mon May 22, 2017 7:23 pm ]
Post subject:  Re: Predict AlphaGo on Future of Go Summit in Wuzhen, China

hyperpape wrote:
I see that there's going to be a stream at the DeepMind site, but will there be any live commentaries? I think I'm too tired tonight, but I'll probably end up staying up to watch the beginning in spite of that.

Michael Redmond and Stephanie Yin are commenting at https://www.youtube.com/watch?v=Z-HL5nppBnM, which I think is the same as the DeepMind stream.

Page 3 of 7 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/