Main
Date: 06 Dec 2006 04:11:31
From: diegoami
Subject: Man vs Machine .- unfair ?
The way matches Man vs Computer are played now is not fair

I think a panel of GMs should play Deep Fritz. Or maybe the GM should
be allowed to move pieces around on the chessboard when playing a
computer.

Now that would make for interesting chess games IMHO. You don't need to
dumb down the computer, you need to give men more tools

Regards
Diego Amicabile





 
Date: 10 Dec 2006 21:29:39
From: MaximRecoil
Subject: Re: Man vs Machine .- unfair ?

Thomas T. Veldhouse wrote:
> Ruud <[email protected]> wrote:
> >
> > Good idea!!
> > The GM should have an analysis-board at his side, on the playing table,
> > when playing engines.
> > And maybe some theory on tablebases as well.
> > That could improve man's chances
> >
>
> Hmm ... unfair?
>
> You sit down and play a human OTB ... you get your scoresheet and your brain.
> You sit down and play a computer OTB ... you should get your scoresheet and
> your brain. The computer gets an openning book, just like humans have an
> opportunity to memorize opening book. Same thing goes for tablebases.

So if we were to play, you get your scoresheet and your brain while I
get my scoresheet, brain and an opening book, endgame tablebases and
historical games all loaded on a laptop with excellent search
functions. I can't be bothered to memorize that stuff, and even if I
could, I would still prefer the reference material to memory, and since
you have already implied that you consider human memory to be the
equivalent of tangible reference material, you shouldn't have any
problems with such a setup. Oh, and this match should be played for
lots and lots of cash.



 
Date: 06 Dec 2006 21:32:13
From: Simon Waters
Subject: Re: Man vs Machine .- unfair ?
> Ruud wrote:
>
> Can we similarly forbid the GMs from using any knowledge they've gained
> from computer analysis?

With most humans one can merely wait five minutes. Works for my brain.

Probably why we have the same threads here every time a man loses to a
machine at chess.


  
Date: 06 Dec 2006 22:28:03
From: Kenneth Sloan
Subject: Re: Man vs Machine .- unfair ?
Simon Waters wrote:
>> Ruud wrote:
>>
>> Can we similarly forbid the GMs from using any knowledge they've gained
>> from computer analysis?
>
> With most humans one can merely wait five minutes. Works for my brain.

Are you a GM?

>
> Probably why we have the same threads here every time a man loses to a
> machine at chess.


--
Kenneth Sloan [email protected]
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/


 
Date: 06 Dec 2006 15:40:14
From: Ruud
Subject: Re: Man vs Machine .- unfair ?

"diegoami" <[email protected] > schreef in bericht
news:[email protected]...
> The way matches Man vs Computer are played now is not fair
>
> I think a panel of GMs should play Deep Fritz. Or maybe the GM should
> be allowed to move pieces around on the chessboard when playing a
> computer.

Good idea!!
The GM should have an analysis-board at his side, on the playing table,
when playing engines.
And maybe some theory on tablebases as well.
That could improve man's chances





  
Date: 06 Dec 2006 15:59:42
From: Thomas T. Veldhouse
Subject: Re: Man vs Machine .- unfair ?
Ruud <[email protected] > wrote:
>
> Good idea!!
> The GM should have an analysis-board at his side, on the playing table,
> when playing engines.
> And maybe some theory on tablebases as well.
> That could improve man's chances
>

Hmm ... unfair?

You sit down and play a human OTB ... you get your scoresheet and your brain.
You sit down and play a computer OTB ... you should get your scoresheet and
your brain. The computer gets an openning book, just like humans have an
opportunity to memorize opening book. Same thing goes for tablebases.

Kramnik lost ... plain and simple ... he simply was not as good as the
computer at the time he played.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0




   
Date: 06 Dec 2006 17:44:16
From: Ruud
Subject: Re: Man vs Machine .- unfair ?
>The computer gets an openning book, just like humans have an
> opportunity to memorize opening book.

Humans have this through years of practice (centuries), millions of games
(billion).
Computers have an openingbook in a few hours, gathering some databases and
editing by a programmer.
Let a computer develope it's own theories, I would say, and forbid it using
human experience. ;-)




    
Date: 06 Dec 2006 12:43:49
From: Kenneth Sloan
Subject: Re: Man vs Machine .- unfair ?
Ruud wrote:
>> The computer gets an openning book, just like humans have an
>> opportunity to memorize opening book.
>
> Humans have this through years of practice (centuries), millions of games
> (billion).
> Computers have an openingbook in a few hours, gathering some databases and
> editing by a programmer.
> Let a computer develope it's own theories, I would say, and forbid it using
> human experience. ;-)
>
>

Can we similarly forbid the GMs from using any knowledge they've gained
from computer analysis?

--
Kenneth Sloan [email protected]
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/


     
Date: 06 Dec 2006 15:04:53
From: Ange1o DePa1ma
Subject: Re: Man vs Machine .- unfair ?
"Kenneth Sloan" <[email protected] > wrote in message
news:[email protected]...
> Ruud wrote:
>>> The computer gets an openning book, just like humans have an
>>> opportunity to memorize opening book.
>>
>> Humans have this through years of practice (centuries), millions of games
>> (billion).
>> Computers have an openingbook in a few hours, gathering some databases
>> and editing by a programmer.
>> Let a computer develope it's own theories, I would say, and forbid it
>> using human experience. ;-)
>
> Can we similarly forbid the GMs from using any knowledge they've gained
> from computer analysis?

We can never completely reconcile differences between how humans and
computers think. Another poster rhetorically asked what the difference was
between storage and processing.

But there is a difference between storage and processing, which I'll
illustrate. Let's say the game of chess has been solved for all moves
leading up to a "one pawn" advantage for both sides. I'm assuming that's
enough for a top GM or top engine to win. And let's say that data file were
stored on two computers, one a standalone machine with a chess engine, and
the other accessible to a human. And let's say a human utilizing this
database and a proxy for the engine (also consulting the tables) sat down to
play a match. Would anyone be interested? I would not because all they'd be
doing is following the thread.

Would such a match be more interesting if only the computer were permitted
to consult the database? Would it be fair? That is essentially what modern
opening theory is all about. Getting an advantage as far into the game as
you possibly can. Ok, instead of move 12, 19, or 25 "book" would end at move
43. Kasparov's "Great Predecessors" might have called it "middlegame theory"
since as some have pointed out GMs know certain lines beyond 25 moves.

Using a full opening book is unfair for two reasons. First of all the human
has got to think about those moves, whereas an entity consulting a "book"
can play them automatically. Remember, even when booked up players think
about which variation they will use. Second, a human is capable of
forgetting, which every player does at some point during a game. A machine
can't forget because it has the book open. If the computer is so brilliant,
if it can see 8 moves ahead on each and every move, why does it need an
opening book anyway?

I would guess that turning off the book, even at move 1, would *at most*
weaken the computer only slightly. Maybe 100 points. If they get into an
even middlegame they'll still destroy almost anyone, regardless. Does anyone
really think that a 2613-rated GM can beat a top engine with the book turned
off? I seriously doubt it.
adp




      
Date: 06 Dec 2006 20:26:45
From: Thomas T. Veldhouse
Subject: Re: Man vs Machine .- unfair ?
Ange1o DePa1ma <[email protected] > wrote:
>

Oh yes ... funny nobody complained until after Kramnik lost. I didn't hear
Kramnik complaing.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0




      
Date: 06 Dec 2006 20:24:13
From: Thomas T. Veldhouse
Subject: Re: Man vs Machine .- unfair ?
Ange1o DePa1ma <[email protected] > wrote:
>
> We can never completely reconcile differences between how humans and
> computers think. Another poster rhetorically asked what the difference was
> between storage and processing.

> But there is a difference between storage and processing, which I'll
> illustrate. Let's say the game of chess has been solved for all moves
> leading up to a "one pawn" advantage for both sides. I'm assuming that's
> enough for a top GM or top engine to win. And let's say that data file were
> stored on two computers, one a standalone machine with a chess engine, and
> the other accessible to a human. And let's say a human utilizing this
> database and a proxy for the engine (also consulting the tables) sat down to
> play a match. Would anyone be interested? I would not because all they'd be
> doing is following the thread.
>

And how do you suppose they would solve the game of chess down to one pawn?
Of course, by computer! A human mind can not do it, because a human skill is
pattern matching and not raw processing power. It is very fair to say that
the computer SHOULD win a game with that knowledge handy. It is not possible
to put that knowledge into a human.

Current programs were built to play chess. They were built to play it well.
They are not given the data as if chess had been solved. They are given the
information they need to play the game. How often do you start like this?

1. a4

I bet not often. Why, because you have in your brain, the knowledge that that
is a poor move. That knowledge is either from learned behaviour of learning
and trusting somebody elses knowledge. This is no different for a computer
using an openning book. Same goes for end game. It is possible for a
computer and a human to play end games perfectly. Humans just take a
different approach to do it.

> Would such a match be more interesting if only the computer were permitted
> to consult the database? Would it be fair? That is essentially what modern
> opening theory is all about. Getting an advantage as far into the game as
> you possibly can. Ok, instead of move 12, 19, or 25 "book" would end at move
> 43. Kasparov's "Great Predecessors" might have called it "middlegame theory"
> since as some have pointed out GMs know certain lines beyond 25 moves.
>

Modern openning books are FAR from a database of chess [solved]. It
represents nothing really different than what a GM knows about openings (i.e.
don't start with 1.a4).

> Using a full opening book is unfair for two reasons. First of all the human
> has got to think about those moves, whereas an entity consulting a "book"
> can play them automatically.

There are many speed chess players that have learned openings and do this very
thing. It is a computers advantage that it can do this faster and with many
more opening positions that most or all people.

> Remember, even when booked up players think
> about which variation they will use. Second, a human is capable of
> forgetting, which every player does at some point during a game. A machine
> can't forget because it has the book open. If the computer is so brilliant,
> if it can see 8 moves ahead on each and every move, why does it need an
> opening book anyway?

It is accumulated knowledge, just like it is for a human or animal. What you
argue is that, since a human doesn't have a photographic memory that can hold
the same information a chess engine has access to, that you should somehow
limit the chess engine to be like a human and be restricted too. Now THAT is
not fair. A human being, to win a game of chess against a computer, will have
to use an attribute or skill uniquely human that a chess engine doesn't have
[pattern matching is a MASSIVE advantage in the human mind that a computer
does not do natively ... that is why it must consider all potential moves and
then determine the best way to evaluate them by pruning the options].

The fact is that raw processing power was enough to beat the human Kramnik.

The only thing that is going to change that is a handicap.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0




       
Date: 06 Dec 2006 21:18:09
From: Ange1o DePa1ma
Subject: Re: Man vs Machine .- unfair ?

"Thomas T. Veldhouse" <[email protected] > wrote

> And how do you suppose they would solve the game of chess down to one
> pawn?

I wasn't saying that. I was saying to create a massive opening book that
contained all the moves needed to get to +1 pawn equivalent, which seems to
be all a computer needs to win against a human.

> Same goes for end game. It is possible for a
> computer and a human to play end games perfectly. Humans just take a
> different approach to do it.

I'm glad you mentioned endgames. Really difficult endgames, with multiple
pieces, cannot really be studied. There is no algorithm for winning like
there is with K+P. Yet databases exist that allow a computer to announce
mate in 83 from such positions. Left to its own resources the computer might
win or it might not. With the tablebases handy even I could "win" such a
position against Kasparov.

Is that chess or is it some other game?




        
Date: 07 Dec 2006 14:04:53
From: Thomas T. Veldhouse
Subject: Re: Man vs Machine .- unfair ?
Ange1o DePa1ma <[email protected] > wrote:
>
> I'm glad you mentioned endgames. Really difficult endgames, with multiple
> pieces, cannot really be studied. There is no algorithm for winning like
> there is with K+P. Yet databases exist that allow a computer to announce
> mate in 83 from such positions. Left to its own resources the computer might
> win or it might not. With the tablebases handy even I could "win" such a
> position against Kasparov.
>

Hmm ... a computer calculated the tablebases in the first place. They are
trivial to create ... just expensive. It is a computers "own" resource, so
you comment about "left to its own resources" really means "if you cripple the
computer so that a human can beat it".

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0




         
Date: 07 Dec 2006 14:57:01
From: David Richerby
Subject: Re: Man vs Machine .- unfair ?
Thomas T. Veldhouse <[email protected] > wrote:
> Ange1o DePa1ma <[email protected]> wrote:
>> I'm glad you mentioned endgames. Really difficult endgames, with
>> multiple pieces, cannot really be studied. There is no algorithm
>> for winning like there is with K+P. Yet databases exist that allow
>> a computer to announce mate in 83 from such positions. Left to its
>> own resources the computer might win or it might not. With the
>> tablebases handy even I could "win" such a position against
>> Kasparov.

Of course, Kasparov would never let you get into such a position
against him so this is kind of moot.


> Hmm ... a computer calculated the tablebases in the first place.
> They are trivial to create ... just expensive. It is a computers
> "own" resource

Yes but in most (all?) cases, it couldn't calculate the tablebase in
the time available in a single game.


> so you comment about "left to its own resources" really means "if
> you cripple the computer so that a human can beat it".

Well, exactly.


Dave.

--
David Richerby Swiss Apple (TM): it's like a tasty
www.chiark.greenend.org.uk/~davidr/ fruit but it's made in Switzerland!


    
Date: 06 Dec 2006 16:55:16
From: Thomas T. Veldhouse
Subject: Re: Man vs Machine .- unfair ?
Ruud <[email protected] > wrote:
>>The computer gets an openning book, just like humans have an
>> opportunity to memorize opening book.
>
> Humans have this through years of practice (centuries), millions of games
> (billion).
> Computers have an openingbook in a few hours, gathering some databases and
> editing by a programmer.
> Let a computer develope it's own theories, I would say, and forbid it using
> human experience. ;-)
>

Kramnik was beat fair and square. You'll live.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0




 
Date:
From: Martin Brown
Subject: Re: Man vs Machine .- unfair ?