Robots and chess: Letter to 'The Times' 8 Dec. 2017

Discuss anything you like about chess related matters in this forum.
Post Reply
Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Paul Habershon » Fri Dec 08, 2017 6:32 pm

Letter in today's 'Times' from Peter West, London SW20: ROBOTS AND CHESS
Sir, I am not surprised that a computer can quickly master chess ('Robot becomes world-beating chess grandmaster in four hours' Dec 7). I enjoyed chess as a child but my performance gradually got worse. I found that as well as using their skill, many opponents were memorising games. Once many games are memorised, a chess player can look for positions that come from a known game, with known winning moves. As memory, rather than skill,
plays such a large part of success in chess, a computer will obviously be able to master the memory side of the game very quickly. We should perhaps reassess our view of chess as a game of skill rather than our view of computers.


I don't know where to start with a reply to this. I'm hoping that the thread will continue in the newspaper tomorrow with replies from reputable and articulate chess brains.

Brian Towers
Posts: 1266
Joined: Tue Nov 18, 2014 7:23 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Brian Towers » Fri Dec 08, 2017 7:26 pm

He is more or less right though, isn't he? It's not whole games that players memorise so much as patterns and short sequences of moves.

Didn't one of the top players recently play a game and afterwards say that he had had the position after 38 moves in his preparation? He took a long time over his moves not because he was calculating but because he was trying to remember his prep. 38 moves isn't a whole game but it is a significant part of one.
Ah, but I was so much older then. I'm younger than that now.

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by NickFaulks » Fri Dec 08, 2017 11:27 pm

Brian Towers wrote:
Fri Dec 08, 2017 7:26 pm
He is more or less right though, isn't he?
Yes he is, except that I might disagree with his assumption that memory is not a skill.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Ian Thompson
Posts: 3551
Joined: Wed Jul 02, 2008 4:31 pm
Location: Awbridge, Hampshire

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Ian Thompson » Sat Dec 09, 2017 1:04 am

Brian Towers wrote:
Fri Dec 08, 2017 7:26 pm
He is more or less right though, isn't he? It's not whole games that players memorise so much as patterns and short sequences of moves.
That's what players do. I assume his letter has been prompted by AlphaZero. We don't know (do we?) how it evaluates positions.

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by NickFaulks » Sat Dec 09, 2017 1:48 am

Ian Thompson wrote:
Sat Dec 09, 2017 1:04 am
We don't know (do we?) how it evaluates positions.
I think the simple version is that it sees how similar positions ( defining "similar" sounds like the tricky bit ) have turned out in previous experiments.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Geoff Chandler
Posts: 3486
Joined: Mon Jul 06, 2009 1:36 pm
Location: Under Cover
Contact:

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Geoff Chandler » Sat Dec 09, 2017 3:22 am

Of course he is wrong.

if he is right then the old Masters, especially Greco were better chess players than todays players.
(what games did he/they memorise to produce their combinations.) and we know that is not the case.

We recall patterns and ideas from other games and have a fair understanding of technique
mixed in with memory and our imaginations.

Who was he playing as a kid that memorised whole games and beat him and soured his outlook on the game.
He possibly kept walking into opening traps, got fed up getting beat, refused to open a chess book and left the game.

Out of the opening computers do not access a database of every known game, they calculate
and number crunch without the baggage humans have. Every middle game position is fresh to them.

The have no memory of ideas or imagination which is a bonus the way they work. A lot of human lose games
because they have an OTB idea, often lifted from a previous game they have seen, which has a flaw in it.
Best to let the other guy have ideas, the counter idea is stronger.

Viktor Korchnoi - Salomon Flohr, USSR Ch, 1955

Korchnoi (to play) gets an idea in an equal position.



26.Rxb7 Rxb7 27. c6

Flohr did not force White into this it was all Korchnoi's idea.

27....Rc7!



White resigned. There are countless examples of players having ideas and resigning a few moves later.

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by NickFaulks » Sat Dec 09, 2017 10:37 am

Geoff Chandler wrote:
Sat Dec 09, 2017 3:22 am
Out of the opening computers do not access a database of every known game, they calculate and number crunch without the baggage humans have. Every middle game position is fresh to them.
That is true for traditional programs, but for Alpha Zero it is precisely false. Without its memory, it has nothing.

While there are more legal games of chess than there are atoms in the universe, it may be that there are only a few billion essentially different chess positions, and it has seen and remembered nearly all of them. At some level it must suffer from Korchnoi-Flohr syndrome, but it has buried that very deep.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

User avatar
JustinHorton
Posts: 10364
Joined: Mon Aug 04, 2008 10:06 am
Location: Somewhere you're not

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by JustinHorton » Sat Dec 09, 2017 11:34 am

Also, pattern recognition may occasionally betray us but by and large it is an enormously important part of our chess strength.
"Do you play chess?"
"Yes, but I prefer a game with a better chance of cheating."

lostontime.blogspot.com

Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Paul Habershon » Sat Dec 09, 2017 3:24 pm

I'm hoping that the thread will continue in the newspaper tomorrow with replies from reputable and articulate chess brains.
As I'd hoped, here is today's letter ('Times' Dec. 9th) Man v machine from Michael Brigden (Bath) grade 179, but I'm old enough to remember him decades ago as 200+ in the Bristol area. The letter echoes some of the views expressed above.

Sir, Peter West (letter Dec.8th) is wrong to suppose that computers' great strength at chess is based on memory. It is their ability to calculate so accurately and quickly that produces their advantage over humans. In terms of general understanding and strategy of the game, areas where memory can play an important part, humans are still far superior.

I will be surprised if any further contributions are published in the new week beginning Dec. 11th.

NickFaulks
Posts: 8453
Joined: Sat Jan 02, 2010 1:28 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by NickFaulks » Sat Dec 09, 2017 5:15 pm

Paul Habershon wrote:
Sat Dec 09, 2017 3:24 pm
I will be surprised if any further contributions are published in the new week beginning Dec. 11th.
It seems strange that these letters assert a view that I think we all ( aside from the Deep Mind team ) would have found completely reasonable until a week ago, but has now been shown to be quite wrong.
Last edited by NickFaulks on Mon Dec 11, 2017 3:01 pm, edited 1 time in total.
If you want a picture of the future, imagine a QR code stamped on a human face — forever.

Paul Habershon
Posts: 550
Joined: Sat Aug 07, 2010 5:51 pm

Re: Robots and chess: Letter to 'The Times' 8 Dec. 2017

Post by Paul Habershon » Mon Dec 11, 2017 2:38 pm

Paul Habershon wrote: ↑Sat Dec 09, 2017 3:24 pm
I will be surprised if any further contributions are published in the new week beginning Dec. 11th.
Unexpectedly chess continues to feature on the Times letters page today, though the thread has broadened. Today's contribution is somewhat tongue-in-cheek (ref. a club of two people) but it is coincidentally relevant to the new 'General Chat' thread about attracting new players. To me chess has the same problem as bridge, also bedevilled by dwindling congress and club attendance (with exceptions, of course) and by its aged demographic. In both games there is a divide between those who see it as a social pastime ('kitchen bridge') and club players (mostly masculine and serious - see letter below, but bridge has many more women) who see it as competitive, though that should not exclude friendliness and sociability. Each game also has also lost participants who play exclusively online. I saw a good solution at the Timosoara (Romania) Chess Club in 1979: a ground floor with a coffee-house atmosphere for noisy skittles and chat, then a quiet room upstairs for matches and serious games. I don't know if it was state funded, but how many chess clubs in the UK could have permanent premises for that sort of set-up?

Anyway here is today's letter from Andy Palmer, Mappowder, Dorset. Sir, I, and all of the Friendly Chess Society of Mappowder in North Dorset (est. November 2017), agree entirely with your correspondent ('Robots and Chess' Letters Dec. 7).

One learns little about the personality of one's opponent when one plays a computer. Such a practice is soulless, solitary and entirely without the fellowship, laughter and personal development which good chess can bring.

If we can personalise chess, perhaps new friendly clubs like ours - our membership is two -might increase. We'd love beginners and young people as it's a lovely sociable game, but it has a reputation of being masculine and serious.

Post Reply