Computer Go

Discuss anything you like about chess related matters in this forum.
Post Reply
Andy Price
Posts: 28
Joined: Tue Apr 03, 2007 12:59 pm

Computer Go

Post by Andy Price » Wed Jan 27, 2016 6:48 pm

A computer has just beaten a Go professional 5-0 in a match.
Our Go playing counterparts may be needing some advice soon
about measures to prevent cheating, although I'm sure the Go
world is much more gentlemanly than our game??

Roger de Coverly
Posts: 17998
Joined: Tue Apr 15, 2008 2:51 pm

Re: Computer Go

Post by Roger de Coverly » Wed Jan 27, 2016 7:01 pm

Andy Price wrote:Our Go playing counterparts may be needing some advice soon
about measures to prevent cheating
The general principle has been in the Laws of Chess for decades. In its current wording:-
11.3 (a) wrote:During play the players are forbidden to use any notes, sources of information or advice, or analyse any game on another chessboard.
Does Go have something similarly worded?

Mobile phones have changed from being a noise hazard to being a cheating hazard.

Nominally 11.3(b) says
11.3(b) wrote:During play, a player is forbidden to have a mobile phone and/or other electronic means of communication in the playing venue. If it is evident that a player brought such a device into the playing venue, he shall lose the game. The opponent shall win.
The rules of a competition may specify a different, less severe, penalty.
The arbiter may require the player to allow his clothes, bags or other items to be inspected, in private. The arbiter or a person authorised by the arbiter shall inspect the player and shall be of the same gender as the player. If a player refuses to cooperate with these obligations, the arbiter shall take measures in accordance with Article 12.9.
The practice is less onerous, recognising that players will have a phone for potential use before and after the game. Some British arbiters will apply a rule to put in in a bag during play, to ensure that it cannot be consulted whilst in play.

MartinCarpenter
Posts: 2433
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Wed Jan 27, 2016 7:34 pm

Rather ignorant but don't think there's any terribly meaningful professional Go in the UK (or Europe), China/Japan etc probably rather different, but goodness knows what the culture is like :)
(Beyond it being very seriously studied as an abstract game.).

Bit sad to have one of the last bastions of human supremacy crumbling rather, but technology marches on....

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Thu Jan 28, 2016 8:31 am

Yes computers can now beat Go professionals, but only because of years of work by the DeepMind company, headed by chess player Demis Hassabis! 8)

http://www.bbc.co.uk/news/technology-35420579

stevencarr

Re: Computer Go

Post by stevencarr » Thu Jan 28, 2016 10:02 am

Andy Price wrote:A computer has just beaten a Go professional 5-0 in a match.
Our Go playing counterparts may be needing some advice soon
about measures to prevent cheating, although I'm sure the Go
world is much more gentlemanly than our game??
It is extraordinary that the European Champion was beaten by a machine.

Note that the computer takes an entirely different approach to playing Go than to playing chess.

MartinCarpenter
Posts: 2433
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Thu Jan 28, 2016 10:50 am

I was just reading the paper describing it, and it really is striking just how different it is!

There's a bit in where they mention winning 85% of their games vs a leading open source program while doing no search at all. Basically equivalent in chess terms to just using the evaluation function to pick the best move. Almost shockingly human really :)
(They do do standard search as well, but they're vastly less dependent on it - the search space is far too large for that sort of technique to work well.).

It would actually be fascinating to see what this sort of approach produced if they turned it on chess. No 'need' for it of course - and it might not work as well anyway - but it'd be so different to modern engines.

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Thu Jan 28, 2016 12:06 pm

MartinCarpenter wrote:I was just reading the paper describing it, and it really is striking just how different it is!
They do do standard search as well, but they're vastly less dependent on it - the search space is far too large for that sort of technique to work well.
Since there are so many reasonable-looking candidate moves in Go, I wonder how human players go about selecting them and how many they consider for further analysis?

Would it be a similar process to a Grandmaster choosing a chess move?

(I don't play Go at all).

John McKenna
Posts: 3632
Joined: Tue May 17, 2011 2:02 pm

Re: Computer Go

Post by John McKenna » Thu Jan 28, 2016 12:23 pm

Here's an American take on it -

http://www.usgo.org/news/2016/01/alphag ... i-advance/
Since there are so many reasonable-looking candidate moves in Go, I wonder how human players go about selecting them and how many they consider for further analysis? Would it be a similar process to a Grandmaster choosing a chess move?
It could be better to go at that from the point of view of checkers (draughts) rather than chess.
To find a for(u)m that accommodates the mess, that is the task of the artist now. (Samuel Beckett)

stevencarr

Re: Computer Go

Post by stevencarr » Thu Jan 28, 2016 12:32 pm

Clive Blackburn wrote:
MartinCarpenter wrote:I was just reading the paper describing it, and it really is striking just how different it is!
They do do standard search as well, but they're vastly less dependent on it - the search space is far too large for that sort of technique to work well.
Since there are so many reasonable-looking candidate moves in Go, I wonder how human players go about selecting them and how many they consider for further analysis?

Would it be a similar process to a Grandmaster choosing a chess move?

(I don't play Go at all).
There are players who just play shapes - what looks like a good shape, without calculation. You can play quite reasonably just playing bamboo joints , hanes, one point jumps, avoiding empty triangles etc etc

I am a really bad Go player though.....

User avatar
David Shepherd
Posts: 848
Joined: Fri Nov 23, 2007 3:46 pm

Re: Computer Go

Post by David Shepherd » Thu Jan 28, 2016 12:46 pm

Looks like a major advance - give the computer two or three stones handicap and it might even come close to beating the very top players in the world :?

MartinCarpenter
Posts: 2433
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Thu Jan 28, 2016 1:07 pm

Well they're going to try it vs the WC :) Doubt the first effort will come off.

I was just thinking about how disturbingly(?) effective this general sort of approach could be if turned on poker/bridge. Typically its been very hard to model humans but there's now a quite vast quantity of data out there in terms of how humans (mis)play those games online......

Doubt if anyone will bother for bridge though.


User avatar
Michael Farthing
Posts: 1787
Joined: Fri Apr 04, 2014 1:28 pm
Location: Morecambe, Europe

Re: Computer Go

Post by Michael Farthing » Fri Jan 29, 2016 3:31 pm

Sadly, I shall not be attending Frodsham this year because of other commitments, but in previous years, though not taking the byes needed to enter the GO, I have boldly entered their room and been treated to between-rounds freely given introductory tuition. So anyone wanting to sample this game would be well-advised to visit Frodsham.

Clive Blackburn

Re: Computer Go

Post by Clive Blackburn » Sat Feb 06, 2016 9:44 am

The DeepMind program will challenge the Go world champion over 5 games, live on YouTube in March.

http://www.wired.co.uk/news/archive/201 ... ive-stream

MartinCarpenter
Posts: 2433
Joined: Tue May 24, 2011 10:58 am

Re: Computer Go

Post by MartinCarpenter » Sat Feb 06, 2016 11:07 am

Expected to lose I think, but I guess Google could always loan it a spare data center or two.....

Post Reply