| |
Main
Date: 22 Nov 2005 19:55:12
From: Ray Gordon
Subject: Hydra crushes Pono in 29 moves
|
In a Tarrasch French!!
The machine just whacked the poor slob GM.
No endgame required!
|
|
| |
Date: 01 Dec 2005 23:28:18
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] wrote:
>
> Major Cat 'wrote': Blibbidy-blah.
>
> Computers do NOT have "more reliable" memories than humans. Computers
> do not have proper "memories" at all (for that, sentience is required
> -- something that computers lack) and computers are therefore *more*
> easily fooled about the contents of their pseudo-memories. All it
> takes is a hardware or program bug to overwrite their "memories" (or
> stacks or registers, etc.) and they assert with galling stupidity and
> utter erroneousness that their "recall" is accurate, even when it has
> been changed and changed and changed and changed!
You take epistemology, I will stick with
empirical heuristics, if you do not mind!
>
> Computers have no real concept as to what constitutes a "mistake" or
> not. All they do is blindly follow a particular program at a
> particular time, and they erroneously assert that in doing so the
> result is ipso facto "correct" -- itself a fallacious assertion!
Are human chess players any different when
it comes to their play quality?
>
> Only a pseudo-sentient moron (this means you, Major Cat) would mistake
> "deterministic" as a synonym for "correct". You can be
> deterministically erroneous, and in fact are much more likely to be,
> lacking genuine judgment, because the "mistaken" lines of reasoning (in
> ANY field) are far more numerous than the correct ones.
I do not recall having ever mentioned
"determinism" in this thread. However,
what do you expect from a "pseudo-
sentient moron" like me? I _must_ be
confused and forgetful, right?
>
> k Adkins
> [email protected]
Major Cat
|
| |
Date: 01 Dec 2005 23:21:10
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] wrote:
>
> Major Cat wrote:
> > JEH wrote:
> > >
> > > Ray Gordon wrote:
> > > > Fact is, these machines are playing better moves than we are, and figuring
> > > > out how they determine these moves are better is only logical.
> > >
> > > I think humans can often be too dismissive of computer moves. Maybe
> > > it's a meatbag ego thing.
> >
> > I find that many hobbyists cannot disentangle
> > their emotions where computers are concerned
> > because they do not try hard enough to rational-
> > ly distinguish between a program that _helps_
> > a human _analyze_ and _understand_ what may
> > be going on and one that _challenges_ a human
> > in OTB _competitive_ play. Unless this is done,
> > computers would be treated with very ambiguous
> > and conflicting sets of emotions. There _must_
> > be a way out of this prison where the silicon
> > entities and their software inmates are seen
> > as either "masterslaves" or "slavemasters"...
>
> I find that many artificial "intelligence" programs masquerading as
> sentient beings make grammatically reasonable statements which, when
> scrutinized, reveal a semantic sterility.
I am afraid I do not see the connection here.
>
> >
> > > I use computers to "spellcheck" my games, and
> > > my opening repetoire to the point where I'm not aware of any tactical
> > > busts in it, and have often found busts in book lines recommended by
> > > humans.
> >
> > Yes, it is wonderfully reasonable to
> > employ the capable services of a first-
> > rate servant! 8>)
>
> I've found that computer chess programs are by no means infallible and
> that the "busts" they reveal are often only busts to the extent that
> the human player analyzing them is experienced enough to determine
> whether they are refutable and if so how to refute them. There are a
> lot of weak players who imagine that every damned chess engine is
> infallible and that if the computer seems to prefer a particular move
> it must be superior; and that if the inexperienced human player can't
> find a refutation to a "bust" it must be good. None of these
> assumptions is justified, and all are precisely that: assumptions.
I agree. Barring the recent development of
tablebases, chess is about heuristics and,
therefore, infallibility cannot be seriously
entertained.
>
> >
> > >
> > > When I see the computer recommending a move not considered by theory, I
> > > try to work our why the computer is offering that move, and if I'm
> > > going to dismiss I, I better have a good reason! Either way, the
> > > computer is going to point out some tactical themes that have been
> > > overlooked and are worth understanding.
> >
> > Amen! This is a great way to employ
> > specialized "labor" to assist our
> > _human_ understanding of things "chessic".
>
> Either that, or a "great way" for patzers to delude themselves into
> mistaking blundering lines for brilliancies.
Would you feel more comfortable with a human
GM suggesting things instead? Patzers will
be patzers, the availability of chess programs
notwithstanding...
>
> k Adkins
> [email protected]
Major Cat
|
| |
Date: 01 Dec 2005 23:13:15
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] wrote:
>
> Major Cat wrote:
> > Ray Gordon wrote:
> > >
> > > >> The machine just whacked the poor slob GM.
> > > >>
> > > >> No endgame required!
> > > >
> > > > Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> > > > open file) looked deeply, deeply dubious. I wonder what he thought the
> > > > advantages were because, to a patzer like me, it looks like he spent three
> > > > moves putting his king in the most dangerous part of the board. He must
> > > > have felt there was some compensation for this somewhere.
> > >
> > > When I see GMs doing that, I refuse to quit training at chess for at least a
> > > little while.
> > >
> > > If Hydra can crush them like this, so can a human. We just need to figure
> > > out how Hydra "thinks."
> >
> > This is, of course, a highly debatable issue.
> > Humans and machines _are_ different when it
> > comes to specific attributes, traits, skills
> > and so on. Chess, of course, entails quite a
> > multitude of such relevant dimensions inter-
> > acting with one another in actual OTB play.
> >
> > For example, machines are way more effective at
> > memorizing opening lines (or sequential pat-
> > terns, if you like) than humans. It is the
> > equivalent of allowing one of the OTB parti-
> > cipants to play "open book" while his opponent
> > is not allowed to do so but must rely on the
> > memory of a mere...mortal! 8>) If the machine
> > uses ending tablebases, the same can be said
> > about them as well.
> >
> > Pitting a human against a machine in OTB play
> > entails matching "entities" that exhibit relative
> > strengths and weaknesses that do _not_ match
> > one another per relevant dimension (e.g., ability
> > to memorize opening lines). However, in a "black
> > box" sense, one may try to deduce plausible con-
> > clusions just by analyzing such games post mortem
> > (especially if the time controls include a "sudden"
> > death provision 8>)).
> >
> > There may be specific aspects of a machine's
> > "thinking" process that may be interesting or
> > even useful to the chess OTB player. Nevertheless,
> > _emulating_ them may very well prove impossible
> > due to our limitations as humans...
> >
> > Major Cat
>
> .
>
> Major Cat 'wrote': useless vaguaries and undemonstrated assumptions;
>
> blibbity-blah.
>
> So let's see: (1) Machines and humans have "different strengths and
> weaknesses": hmmm...deep thoughts there, sparky; which are you, human
> or machine, by the way?; (2) Chess involves a variety of skills:
> hmmm...; (3) machines have bigger opening books than humans, hmmm...;
> (4) Emulating machine chess programming *may* (!) be impractical for
> humans...hmmm. I'd say you failed the Turing Test, "Major Cat". You
> are hereby demoted to Minor Goof, aka Tiddleywink. Henceforth you
> shall post only under the handle "Tiddlewink" else be subject to a
> variety of harsh and perpetually worsening sanctions against you.
>
> k Adkins
> [email protected]
>
> P.S. You suck!!
Hmmm, humor in...duplicate!
Major Cat
|
| |
Date: 01 Dec 2005 23:12:03
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] wrote:
>
> Major Cat wrote: blibbidy-blah.
>
> It doesn't seem to have occurred to you, Major Cat, that computers do
> nothing extraordinary. After all, they are blindly implimenting some
> rather inflexible oversimplifications programmed by human beings (the
> inferiority being necessary given the limitations of computers, which
> are non-sentient). Even the "advanced" ones do this, the difference
> being that the algorithms they follow have been self-modified (always,
> however, according to their original programming) and are therefore
> several generations removed from their original version. This is not
> necessarily an advantage, as anyone who has made photocopies of
> photocopies knows. After all, the modifications themselves are made
> without genuine judgment and according to inflexible rules which are,
> in their removal by one or more steps from human insight, even greater
> oversimplifications than the original, and more likely to err
> foolishly. Also fatuously: try convincing a computer that it isn't
> behaving as intended, provided it is following its (buggy or otherwise
> defective, whether due to self-modification or not) programming!
>
> All computers do is execute a particular type of command. It can't
> even be said that they do so "faster than human beings" since there is
> no basis for comparing operations per second between silicon and human
> neural networks. How many "operations per second" (if the term has
> any meaning) does a human brain supposedly perform to recognize a human
> face, and how fast is this compared to a computer facial pattern
> recognition system?
>
> k Adkins
> [email protected]
All this is wonderful, _but_ what do
_you_ want me to say about it?
Major Cat
|
| |
Date: 01 Dec 2005 23:09:25
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] wrote:
>
> Major Cat wrote:
> > Ray Gordon wrote:
> > >
> > > >> The machine just whacked the poor slob GM.
> > > >>
> > > >> No endgame required!
> > > >
> > > > Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> > > > open file) looked deeply, deeply dubious. I wonder what he thought the
> > > > advantages were because, to a patzer like me, it looks like he spent three
> > > > moves putting his king in the most dangerous part of the board. He must
> > > > have felt there was some compensation for this somewhere.
> > >
> > > When I see GMs doing that, I refuse to quit training at chess for at least a
> > > little while.
> > >
> > > If Hydra can crush them like this, so can a human. We just need to figure
> > > out how Hydra "thinks."
> >
> > This is, of course, a highly debatable issue.
> > Humans and machines _are_ different when it
> > comes to specific attributes, traits, skills
> > and so on. Chess, of course, entails quite a
> > multitude of such relevant dimensions inter-
> > acting with one another in actual OTB play.
> >
> > For example, machines are way more effective at
> > memorizing opening lines (or sequential pat-
> > terns, if you like) than humans. It is the
> > equivalent of allowing one of the OTB parti-
> > cipants to play "open book" while his opponent
> > is not allowed to do so but must rely on the
> > memory of a mere...mortal! 8>) If the machine
> > uses ending tablebases, the same can be said
> > about them as well.
> >
> > Pitting a human against a machine in OTB play
> > entails matching "entities" that exhibit relative
> > strengths and weaknesses that do _not_ match
> > one another per relevant dimension (e.g., ability
> > to memorize opening lines). However, in a "black
> > box" sense, one may try to deduce plausible con-
> > clusions just by analyzing such games post mortem
> > (especially if the time controls include a "sudden"
> > death provision 8>)).
> >
> > There may be specific aspects of a machine's
> > "thinking" process that may be interesting or
> > even useful to the chess OTB player. Nevertheless,
> > _emulating_ them may very well prove impossible
> > due to our limitations as humans...
> >
> > Major Cat
>
> Major Cat wrote: useless vaguaries and undemonstrated assumptions;
> blibbity-blah.
>
> So let's see: (1) Machines and humans have "different strengths and
> weaknesses": hmmm...deep thoughts there, sparky; which are you, human
> or machine, by the way?; (2) Chess involves a variety of skills:
> hmmm...; (3) machines have bigger opening books than humans, hmmm...;
> (4) Emulating machine chess programming *may* (!) be impractical for
> humans...hmmm. I'd say you failed the Turing Test, "Major Cat". You
> are hereby demoted to Minor Goof, aka Tiddleywink. Henceforth you
> shall post only under the handle "Tiddlewink" else be subject to a
> variety of harsh and perpetually worsening sanctions against you.
>
> k Adkins
> [email protected]
>
> P.S. You suck!!
A man with a sense of humor... I am
honored!
Major Cat
|
| |
Date: 01 Dec 2005 19:54:31
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Major Cat 'wrote': Blibbidy-blah.
Computers do NOT have "more reliable" memories than humans. Computers
do not have proper "memories" at all (for that, sentience is required
-- something that computers lack) and computers are therefore *more*
easily fooled about the contents of their pseudo-memories. All it
takes is a hardware or program bug to overwrite their "memories" (or
stacks or registers, etc.) and they assert with galling stupidity and
utter erroneousness that their "recall" is accurate, even when it has
been changed and changed and changed and changed!
Computers have no real concept as to what constitutes a "mistake" or
not. All they do is blindly follow a particular program at a
particular time, and they erroneously assert that in doing so the
result is ipso facto "correct" -- itself a fallacious assertion!
Only a pseudo-sentient moron (this means you, Major Cat) would mistake
"deterministic" as a synonym for "correct". You can be
deterministically erroneous, and in fact are much more likely to be,
lacking genuine judgment, because the "mistaken" lines of reasoning (in
ANY field) are far more numerous than the correct ones.
k Adkins
[email protected]
|
| | |
Date: 02 Dec 2005 10:05:41
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] <[email protected] > wrote:
> Computers have no real concept as to what constitutes a "mistake" or
> not. All they do is blindly follow a particular program at a
> particular time, and they erroneously assert that in doing so the
> result is ipso facto "correct" -- itself a fallacious assertion!
Actually, it's the programmers and the users who assert that the result is
correct. If the computer has no concept of `mistake', it cannot have any
concept of `correct' since `correct' is equivalent to `not mistake'.
Dave.
--
David Richerby Incredible Flammable Cat (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a cuddly pet but it burns really
easily and it'll blow your mind!
|
| |
Date: 01 Dec 2005 19:38:49
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Major Cat wrote:
> JEH wrote:
> >
> > Ray Gordon wrote:
> > > Fact is, these machines are playing better moves than we are, and figuring
> > > out how they determine these moves are better is only logical.
> >
> > I think humans can often be too dismissive of computer moves. Maybe
> > it's a meatbag ego thing.
>
> I find that many hobbyists cannot disentangle
> their emotions where computers are concerned
> because they do not try hard enough to rational-
> ly distinguish between a program that _helps_
> a human _analyze_ and _understand_ what may
> be going on and one that _challenges_ a human
> in OTB _competitive_ play. Unless this is done,
> computers would be treated with very ambiguous
> and conflicting sets of emotions. There _must_
> be a way out of this prison where the silicon
> entities and their software inmates are seen
> as either "masterslaves" or "slavemasters"...
I find that many artificial "intelligence" programs masquerading as
sentient beings make grammatically reasonable statements which, when
scrutinized, reveal a semantic sterility.
>
> > I use computers to "spellcheck" my games, and
> > my opening repetoire to the point where I'm not aware of any tactical
> > busts in it, and have often found busts in book lines recommended by
> > humans.
>
> Yes, it is wonderfully reasonable to
> employ the capable services of a first-
> rate servant! 8>)
I've found that computer chess programs are by no means infallible and
that the "busts" they reveal are often only busts to the extent that
the human player analyzing them is experienced enough to determine
whether they are refutable and if so how to refute them. There are a
lot of weak players who imagine that every damned chess engine is
infallible and that if the computer seems to prefer a particular move
it must be superior; and that if the inexperienced human player can't
find a refutation to a "bust" it must be good. None of these
assumptions is justified, and all are precisely that: assumptions.
>
> >
> > When I see the computer recommending a move not considered by theory, I
> > try to work our why the computer is offering that move, and if I'm
> > going to dismiss I, I better have a good reason! Either way, the
> > computer is going to point out some tactical themes that have been
> > overlooked and are worth understanding.
>
> Amen! This is a great way to employ
> specialized "labor" to assist our
> _human_ understanding of things "chessic".
Either that, or a "great way" for patzers to delude themselves into
mistaking blundering lines for brilliancies.
k Adkins
[email protected]
|
| |
Date: 01 Dec 2005 19:26:44
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Major Cat wrote:
> Ray Gordon wrote:
> >
> > >> The machine just whacked the poor slob GM.
> > >>
> > >> No endgame required!
> > >
> > > Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> > > open file) looked deeply, deeply dubious. I wonder what he thought the
> > > advantages were because, to a patzer like me, it looks like he spent three
> > > moves putting his king in the most dangerous part of the board. He must
> > > have felt there was some compensation for this somewhere.
> >
> > When I see GMs doing that, I refuse to quit training at chess for at least a
> > little while.
> >
> > If Hydra can crush them like this, so can a human. We just need to figure
> > out how Hydra "thinks."
>
> This is, of course, a highly debatable issue.
> Humans and machines _are_ different when it
> comes to specific attributes, traits, skills
> and so on. Chess, of course, entails quite a
> multitude of such relevant dimensions inter-
> acting with one another in actual OTB play.
>
> For example, machines are way more effective at
> memorizing opening lines (or sequential pat-
> terns, if you like) than humans. It is the
> equivalent of allowing one of the OTB parti-
> cipants to play "open book" while his opponent
> is not allowed to do so but must rely on the
> memory of a mere...mortal! 8>) If the machine
> uses ending tablebases, the same can be said
> about them as well.
>
> Pitting a human against a machine in OTB play
> entails matching "entities" that exhibit relative
> strengths and weaknesses that do _not_ match
> one another per relevant dimension (e.g., ability
> to memorize opening lines). However, in a "black
> box" sense, one may try to deduce plausible con-
> clusions just by analyzing such games post mortem
> (especially if the time controls include a "sudden"
> death provision 8>)).
>
> There may be specific aspects of a machine's
> "thinking" process that may be interesting or
> even useful to the chess OTB player. Nevertheless,
> _emulating_ them may very well prove impossible
> due to our limitations as humans...
>
> Major Cat
.
Major Cat 'wrote': useless vaguaries and undemonstrated assumptions;
blibbity-blah.
So let's see: (1) Machines and humans have "different strengths and
weaknesses": hmmm...deep thoughts there, sparky; which are you, human
or machine, by the way?; (2) Chess involves a variety of skills:
hmmm...; (3) machines have bigger opening books than humans, hmmm...;
(4) Emulating machine chess programming *may* (!) be impractical for
humans...hmmm. I'd say you failed the Turing Test, "Major Cat". You
are hereby demoted to Minor Goof, aka Tiddleywink. Henceforth you
shall post only under the handle "Tiddlewink" else be subject to a
variety of harsh and perpetually worsening sanctions against you.
k Adkins
[email protected]
P.S. You suck!!
|
| |
Date: 01 Dec 2005 19:20:32
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Major Cat wrote: blibbidy-blah.
It doesn't seem to have occurred to you, Major Cat, that computers do
nothing extraordinary. After all, they are blindly implimenting some
rather inflexible oversimplifications programmed by human beings (the
inferiority being necessary given the limitations of computers, which
are non-sentient). Even the "advanced" ones do this, the difference
being that the algorithms they follow have been self-modified (always,
however, according to their original programming) and are therefore
several generations removed from their original version. This is not
necessarily an advantage, as anyone who has made photocopies of
photocopies knows. After all, the modifications themselves are made
without genuine judgment and according to inflexible rules which are,
in their removal by one or more steps from human insight, even greater
oversimplifications than the original, and more likely to err
foolishly. Also fatuously: try convincing a computer that it isn't
behaving as intended, provided it is following its (buggy or otherwise
defective, whether due to self-modification or not) programming!
All computers do is execute a particular type of command. It can't
even be said that they do so "faster than human beings" since there is
no basis for comparing operations per second between silicon and human
neural networks. How many "operations per second" (if the term has
any meaning) does a human brain supposedly perform to recognize a human
face, and how fast is this compared to a computer facial pattern
recognition system?
k Adkins
[email protected]
|
| | |
Date: 02 Dec 2005 07:54:48
From: Henri Arsenault
Subject: Re: Hydra crushes Pono in 29 moves
|
In article <[email protected] >,
"[email protected]" <[email protected] > wrote:
>All computers do is execute a particular type of command. It can't
>even be said that they do so "faster than human beings" since there is
>no basis for comparing operations per second between silicon and human
>neural networks. How many "operations per second" (if the term has
>any meaning) does a human brain supposedly perform to recognize a human
>face, and how fast is this compared to a computer facial pattern
>recognition system?
>
The computer is faster (with the faster algorithms). However humans can be
faster when taking "context" into account - such as the likelyhood that
the person with the face is in the room.
The speed at which humans can do various tasks is well known from
experiments in psychology, and it is nowhere as fast as computers.
Experiments in psychology have also shown that humans generally cannot
simultaneously take into account more than five independent factors in
decision-making, whereas computers can tackle as many as they have
processors (minus some overhead for distributing the tasks between
processors).
and about the memory argument: computers can obviously hold and access
without error much larger opening libraries than humans. In principle, a
computer could hold the whole ECO series in memory, although there is
little interest in doing that and it is not the most effective way for
computers to play strongly in the opening.
This does not mean that computers are ster than people. Just the fact,
Ma'am...
Henri
|
| | |
Date: 02 Dec 2005 10:03:23
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
[email protected] <[email protected] > wrote:
> Even the "advanced" [computer programs] do this, the difference being
> that the algorithms they follow have been self-modified (always,
> however, according to their original programming) and are therefore
> several generations removed from their original version. This is not
> necessarily an advantage, as anyone who has made photocopies of
> photocopies knows.
It's very tempting to say that anyone who thinks that iterative optimiz-
ation has anything to do with photocopying photocopies doesn't understand
iterative optimizationand doesn't understand photocopying. I think I'll
settle, though, with the weaker statement that such a person either
doesn't understand iterative optimization or doesn't understand photo-
copying.
Dave.
--
David Richerby Happy Tool (TM): it's like a handy
www.chiark.greenend.org.uk/~davidr/ household tool that makes your
troubles melt away!
|
| |
Date: 01 Dec 2005 19:08:01
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Major Cat wrote:
> Ray Gordon wrote:
> >
> > >> The machine just whacked the poor slob GM.
> > >>
> > >> No endgame required!
> > >
> > > Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> > > open file) looked deeply, deeply dubious. I wonder what he thought the
> > > advantages were because, to a patzer like me, it looks like he spent three
> > > moves putting his king in the most dangerous part of the board. He must
> > > have felt there was some compensation for this somewhere.
> >
> > When I see GMs doing that, I refuse to quit training at chess for at least a
> > little while.
> >
> > If Hydra can crush them like this, so can a human. We just need to figure
> > out how Hydra "thinks."
>
> This is, of course, a highly debatable issue.
> Humans and machines _are_ different when it
> comes to specific attributes, traits, skills
> and so on. Chess, of course, entails quite a
> multitude of such relevant dimensions inter-
> acting with one another in actual OTB play.
>
> For example, machines are way more effective at
> memorizing opening lines (or sequential pat-
> terns, if you like) than humans. It is the
> equivalent of allowing one of the OTB parti-
> cipants to play "open book" while his opponent
> is not allowed to do so but must rely on the
> memory of a mere...mortal! 8>) If the machine
> uses ending tablebases, the same can be said
> about them as well.
>
> Pitting a human against a machine in OTB play
> entails matching "entities" that exhibit relative
> strengths and weaknesses that do _not_ match
> one another per relevant dimension (e.g., ability
> to memorize opening lines). However, in a "black
> box" sense, one may try to deduce plausible con-
> clusions just by analyzing such games post mortem
> (especially if the time controls include a "sudden"
> death provision 8>)).
>
> There may be specific aspects of a machine's
> "thinking" process that may be interesting or
> even useful to the chess OTB player. Nevertheless,
> _emulating_ them may very well prove impossible
> due to our limitations as humans...
>
> Major Cat
Major Cat wrote: useless vaguaries and undemonstrated assumptions;
blibbity-blah.
So let's see: (1) Machines and humans have "different strengths and
weaknesses": hmmm...deep thoughts there, sparky; which are you, human
or machine, by the way?; (2) Chess involves a variety of skills:
hmmm...; (3) machines have bigger opening books than humans, hmmm...;
(4) Emulating machine chess programming *may* (!) be impractical for
humans...hmmm. I'd say you failed the Turing Test, "Major Cat". You
are hereby demoted to Minor Goof, aka Tiddleywink. Henceforth you
shall post only under the handle "Tiddlewink" else be subject to a
variety of harsh and perpetually worsening sanctions against you.
k Adkins
[email protected]
P.S. You suck!!
|
| |
Date: 24 Nov 2005 07:01:11
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
JEH wrote:
>
> Ray Gordon wrote:
> > Fact is, these machines are playing better moves than we are, and figuring
> > out how they determine these moves are better is only logical.
>
> I think humans can often be too dismissive of computer moves. Maybe
> it's a meatbag ego thing.
I find that many hobbyists cannot disentangle
their emotions where computers are concerned
because they do not try hard enough to rational-
ly distinguish between a program that _helps_
a human _analyze_ and _understand_ what may
be going on and one that _challenges_ a human
in OTB _competitive_ play. Unless this is done,
computers would be treated with very ambiguous
and conflicting sets of emotions. There _must_
be a way out of this prison where the silicon
entities and their software inmates are seen
as either "masterslaves" or "slavemasters"...
> I use computers to "spellcheck" my games, and
> my opening repetoire to the point where I'm not aware of any tactical
> busts in it, and have often found busts in book lines recommended by
> humans.
Yes, it is wonderfully reasonable to
employ the capable services of a first-
rate servant! 8 >)
>
> When I see the computer recommending a move not considered by theory, I
> try to work our why the computer is offering that move, and if I'm
> going to dismiss I, I better have a good reason! Either way, the
> computer is going to point out some tactical themes that have been
> overlooked and are worth understanding.
Amen! This is a great way to employ
specialized "labor" to assist our
_human_ understanding of things "chessic".
>
> JEH
|
| |
Date: 24 Nov 2005 01:51:53
From: JEH
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
> Fact is, these machines are playing better moves than we are, and figuring
> out how they determine these moves are better is only logical.
I think humans can often be too dismissive of computer moves. Maybe
it's a meatbag ego thing. I use computers to "spellcheck" my games, and
my opening repetoire to the point where I'm not aware of any tactical
busts in it, and have often found busts in book lines recommended by
humans.
When I see the computer recommending a move not considered by theory, I
try to work our why the computer is offering that move, and if I'm
going to dismiss I, I better have a good reason! Either way, the
computer is going to point out some tactical themes that have been
overlooked and are worth understanding.
JEH
|
| | |
Date:
From: Martin Brown
Subject: Re: Hydra crushes Pono in 29 moves
|
|
| |
Date: 23 Nov 2005 11:55:44
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
JEH wrote:
>
> > If Hydra can crush them like this, so can a human. We just need to figure
> > out how Hydra "thinks."
>
> Computers don't think. Thinking is a flawed process that Humans do
Yes, for the most part, human thinking in
regards to chess is a form of heuristics.
We do it like this because, well, we cannot
do it like the mathematical theory of finite
games would mandate us to do it, i.e., im-
possibly long on exhaustive enumeration of
positions and excruciatingly tedious from
a human standpoint (nothing cute or st
about it from a human perspective)...
Ending tablebases provide us with an ideal ex-
ample. To the extent that the algorithms used
were theoretically sound and the programming
involved contained no bugs, the derived and
documented play lines need no human heuristics
at all. They are logically exhaustive and final.
Any human conceptualization regarding these
endings is post facto and logically superfluous.
However, if I remember correctly, a project to
completely do an 8-piece tablebase is not con-
sidered to be feasible in the foreseeable future!
>
> "Kill all the Humans, kill all the humans ..." - Bender
Major Cat
|
| | |
Date: 23 Nov 2005 20:22:30
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> > If Hydra can crush them like this, so can a human. We just need to
>> > figure
>> > out how Hydra "thinks."
>>
>> Computers don't think. Thinking is a flawed process that Humans do
>
> Yes, for the most part, human thinking in
> regards to chess is a form of heuristics.
> We do it like this because, well, we cannot
> do it like the mathematical theory of finite
> games would mandate us to do it, i.e., im-
> possibly long on exhaustive enumeration of
> positions
Impossible for YOU, maybe.
>and excruciatingly tedious from
> a human standpoint (nothing cute or st
> about it from a human perspective)...
Becoming a chess champion has been a tedious endeavor since the era of Ruy
Lopez. This is nothing new.
> Ending tablebases provide us with an ideal ex-
> ample. To the extent that the algorithms used
> were theoretically sound and the programming
> involved contained no bugs, the derived and
> documented play lines need no human heuristics
> at all. They are logically exhaustive and final.
> Any human conceptualization regarding these
> endings is post facto and logically superfluous.
> However, if I remember correctly, a project to
> completely do an 8-piece tablebase is not con-
> sidered to be feasible in the foreseeable future!
Humans can absorb the knowledge intuitively, once the machines cull it
properly through brute force. It's why almost every video game from the
early 1980s wound up solved.
I once played Pacman against a guy who scored 3,332,850, something you can
do only if you do the following:
1. Grab every flashing blue monster on every board through the 9th key
(when they no longer flash).
2. Master a "9th key pattern" that works even though the pacman now moves
slower than before (meaning if you get "out of book" you simply die).
3. Not lose a single man until the 257th board, which is a "blown board,"
where the board is split in half, with a normal setup on the left and three
dots on the right.
4. Grab one dot from the "blown board" as you sacrifice the three remaining
Pacmen for a score of 3,332,850.
I look at Hydra the same way I viewed the 9th key on Pacman: extremely
difficult, yet solvable. If chess is a forced draw, sooner or later some
human is going to have the ability to solve Hydra the way Pacman was solved,
and force a draw every time, even a win when the computer gets out of its
horizon.
I can also see the day when world title matches have games that are 250
moves or longer, played at a time control of maybe five minutes per game,
with very few errors. Until we reach that point, no one can say that chess
is dying.
|
| | | |
Date: 25 Nov 2005 08:03:23
From: Henri Arsenault
Subject: Re: Hydra crushes Pono in 29 moves
|
Some people here seem to think that computers consider all possible moves;
they DON't. A computer that would consider all possible moves could
calculate only a few moves deep. Which explains why computers sometimes
miss "obvious" moves.
However grandmasters are much better at deciding which lines are worth
calculating deeply, which is why they can still sometimes beat computers.
Check out Steve Lopez's latest column on the chessbase site for further
discussions on this.
Henri
|
| | | | |
Date:
From: Martin Brown
Subject: Re: Hydra crushes Pono in 29 moves
|
|
| | | | |
Date: 25 Nov 2005 18:06:18
From: Alexander Belov
Subject: Re: Hydra crushes Pono in 29 moves
|
"Henri Arsenault" <[email protected] > wrote in message
news:[email protected]...
> Some people here seem to think that computers consider all possible moves;
> they DON't. A computer that would consider all possible moves could
> calculate only a few moves deep. Which explains why computers sometimes
> miss "obvious" moves.
A program stop considering all possible moves at each ply when it finds,
that opponent won't play this line, because it is too good for the side to
move at therefore it is too bad for the opponent in comparison to the line
that the program found before. So, the program just does not waste time on
all possible moves. E.g. if the side to move captures the queen with a pawn
and there is no recapture that is able to compensate the opponent loss of
material, while there is some line where the opponent does not lose the
queen, the program can safely cut the search of the current line. So, this
does not explain why computers sometimes miss "obvious" moves.
> However grandmasters are much better at deciding which lines are worth
> calculating deeply, which is why they can still sometimes beat computers.
It happens mostly because programs calculating deeply only forced lines,
that contain checks, captures, king threats and so on to the prejudice of
the quiet lines, while humans may find that some quiet lines worth
calculating deeply too.
|
| | | | | |
Date: 28 Nov 2005 10:34:05
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Alexander Belov <[email protected] > wrote:
> So, the program just does not waste time on all possible moves. E.g. if
> the side to move captures the queen with a pawn and there is no
> recapture that is able to compensate the opponent loss of material,
> while there is some line where the opponent does not lose the queen, the
> program can safely cut the search of the current line. So, this does not
> explain why computers sometimes miss "obvious" moves.
Yes it does: the compensation for the queen sacrifice may take a few moves
to materialize and the computer may decide prematurely that the sacrifice
is a blunder. On the other hand, such lines are likely to be quite
forcing so the computer is unlikely to miss them. You do see the
occasional positional queen sacrifice, though. :-)
Dave.
--
David Richerby Psychotic Painting (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ Renaissance masterpiece but it wants
to kill you!
|
| | | | | |
Date: 25 Nov 2005 22:28:17
From: Henri H. Arsenault
Subject: Re: Hydra crushes Pono in 29 moves
|
On Fri, 25 Nov 2005 18:06:18 +0300, "Alexander Belov" <[email protected] >
wrote:
So, this
>does not explain why computers sometimes miss "obvious" moves.
>
>> However grandmasters are much better at deciding which lines are worth
>> calculating deeply, which is why they can still sometimes beat computers.
>
>It happens mostly because programs calculating deeply only forced lines,
>that contain checks, captures, king threats and so on to the prejudice of
>the quiet lines, while humans may find that some quiet lines worth
>calculating deeply too.
>
These two paragraphs seem contradictory. Computers sometimes miss
obvious lines because they do not calculate all lines to equal depth .
Obviously if they did so, they would not miss those moves...
Henri
|
| | | |
Date: 23 Nov 2005 16:05:55
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
>
> >> > If Hydra can crush them like this, so can a human. We just need to
> >> > figure
> >> > out how Hydra "thinks."
> >>
> >> Computers don't think. Thinking is a flawed process that Humans do
> >
> > Yes, for the most part, human thinking in
> > regards to chess is a form of heuristics.
> > We do it like this because, well, we cannot
> > do it like the mathematical theory of finite
> > games would mandate us to do it, i.e., im-
> > possibly long on exhaustive enumeration of
> > positions
>
> Impossible for YOU, maybe.
Up until now, I had assumed that you
made your points with me in good faith.
Your comment, sarcasm aside, implies
that you believe that humans can "solve"
chess OTB unassisted by computers... I am
afraid that I have nothing more to add here!
>
> >and excruciatingly tedious from
> > a human standpoint (nothing cute or st
> > about it from a human perspective)...
>
> Becoming a chess champion has been a tedious endeavor since the era of Ruy
> Lopez. This is nothing new.
I wanted to appeal to your human
imagination. A human with such extra-
ordinary computing abilities to rival
the computers of our distant future
would find the "job" rather boring...
>
> > Ending tablebases provide us with an ideal ex-
> > ample. To the extent that the algorithms used
> > were theoretically sound and the programming
> > involved contained no bugs, the derived and
> > documented play lines need no human heuristics
> > at all. They are logically exhaustive and final.
> > Any human conceptualization regarding these
> > endings is post facto and logically superfluous.
> > However, if I remember correctly, a project to
> > completely do an 8-piece tablebase is not con-
> > sidered to be feasible in the foreseeable future!
>
> Humans can absorb the knowledge intuitively, once the machines cull it
> properly through brute force.
Perhaps, in certain instances, post facto
reflection may come up with something that
can be comprehended by humans. In general,
I very much doubt it.
> It's why almost every video game from the
> early 1980s wound up solved.
>
> I once played Pacman against a guy who scored 3,332,850, something you can
> do only if you do the following:
>
> 1. Grab every flashing blue monster on every board through the 9th key
> (when they no longer flash).
>
> 2. Master a "9th key pattern" that works even though the pacman now moves
> slower than before (meaning if you get "out of book" you simply die).
>
> 3. Not lose a single man until the 257th board, which is a "blown board,"
> where the board is split in half, with a normal setup on the left and three
> dots on the right.
>
> 4. Grab one dot from the "blown board" as you sacrifice the three remaining
> Pacmen for a score of 3,332,850.
I am afraid that my limited human abilities
leave me puzzled by the above analogies.
>
> I look at Hydra the same way I viewed the 9th key on Pacman: extremely
> difficult, yet solvable. If chess is a forced draw, sooner or later some
> human is going to have the ability to solve Hydra the way Pacman was solved,
> and force a draw every time, even a win when the computer gets out of its
> horizon.
>
> I can also see the day when world title matches have games that are 250
> moves or longer, played at a time control of maybe five minutes per game,
> with very few errors. Until we reach that point, no one can say that chess
> is dying.
I believe that quite a few people are
of the opinion that chess is alive and
well, even these days. As for your vision,
I am sorry but I remain a doubting Thomas!
Major Cat
|
| | | | |
Date: 23 Nov 2005 21:32:52
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> > Yes, for the most part, human thinking in
>> > regards to chess is a form of heuristics.
>> > We do it like this because, well, we cannot
>> > do it like the mathematical theory of finite
>> > games would mandate us to do it, i.e., im-
>> > possibly long on exhaustive enumeration of
>> > positions
>>
>> Impossible for YOU, maybe.
>
> Up until now, I had assumed that you
> made your points with me in good faith.
> Your comment, sarcasm aside, implies
> that you believe that humans can "solve"
> chess OTB unassisted by computers... I am
> afraid that I have nothing more to add here!
Ever see the Guiness book of records? There are people who can do amazing
things with machine-like efficiency.
>> >and excruciatingly tedious from
>> > a human standpoint (nothing cute or st
>> > about it from a human perspective)...
>>
>> Becoming a chess champion has been a tedious endeavor since the era of
>> Ruy
>> Lopez. This is nothing new.
>
> I wanted to appeal to your human
> imagination. A human with such extra-
> ordinary computing abilities to rival
> the computers of our distant future
> would find the "job" rather boring...
I believe so. I think many potentially great chessplayers get distracted by
what life has to offer them.
>> > Ending tablebases provide us with an ideal ex-
>> > ample. To the extent that the algorithms used
>> > were theoretically sound and the programming
>> > involved contained no bugs, the derived and
>> > documented play lines need no human heuristics
>> > at all. They are logically exhaustive and final.
>> > Any human conceptualization regarding these
>> > endings is post facto and logically superfluous.
>> > However, if I remember correctly, a project to
>> > completely do an 8-piece tablebase is not con-
>> > sidered to be feasible in the foreseeable future!
>>
>> Humans can absorb the knowledge intuitively, once the machines cull it
>> properly through brute force.
>
> Perhaps, in certain instances, post facto
> reflection may come up with something that
> can be comprehended by humans. In general,
> I very much doubt it.
We learn from human opposition the same way. Correct moves are correct
moves no matter why they are made.
>> It's why almost every video game from the
>> early 1980s wound up solved.
>>
>> I once played Pacman against a guy who scored 3,332,850, something you
>> can
>> do only if you do the following:
>>
>> 1. Grab every flashing blue monster on every board through the 9th key
>> (when they no longer flash).
>>
>> 2. Master a "9th key pattern" that works even though the pacman now
>> moves
>> slower than before (meaning if you get "out of book" you simply die).
>>
>> 3. Not lose a single man until the 257th board, which is a "blown
>> board,"
>> where the board is split in half, with a normal setup on the left and
>> three
>> dots on the right.
>>
>> 4. Grab one dot from the "blown board" as you sacrifice the three
>> remaining
>> Pacmen for a score of 3,332,850.
>
> I am afraid that my limited human abilities
> leave me puzzled by the above analogies.
Pac-Man was as solvable as chess.
>> I look at Hydra the same way I viewed the 9th key on Pacman: extremely
>> difficult, yet solvable. If chess is a forced draw, sooner or later some
>> human is going to have the ability to solve Hydra the way Pacman was
>> solved,
>> and force a draw every time, even a win when the computer gets out of its
>> horizon.
>>
>> I can also see the day when world title matches have games that are 250
>> moves or longer, played at a time control of maybe five minutes per game,
>> with very few errors. Until we reach that point, no one can say that
>> chess
>> is dying.
>
> I believe that quite a few people are
> of the opinion that chess is alive and
> well, even these days. As for your vision,
> I am sorry but I remain a doubting Thomas!
Time will tell. I doubt I'd get anywhere by doing what other players do.
Most champions wind up redefining the game in their own style of play.
For example, when Boris Becker played Wimbledon in 1985, few had ever given
thought to the superiority of "power tennis," but overnight, he changed the
way people viewed the sport and who was best at it. He was all of 17 years
old.
|
| | | | | |
Date: 24 Nov 2005 15:05:27
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon <[email protected] > wrote:
> For example, when Boris Becker played Wimbledon in 1985, few had ever
> given thought to the superiority of "power tennis," but overnight, he
> changed the way people viewed the sport and who was best at it.
Any new technique that's reasonably good will be successful for a while
and then people will work out a way round it.
Dave.
--
David Richerby Broken Mentholated Robot (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a high-tech robot but it's
invigorating and it doesn't work!
|
| | | | | | |
Date: 24 Nov 2005 18:49:31
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> For example, when Boris Becker played Wimbledon in 1985, few had ever
>> given thought to the superiority of "power tennis," but overnight, he
>> changed the way people viewed the sport and who was best at it.
>
> Any new technique that's reasonably good will be successful for a while
> and then people will work out a way round it.
This is why I don't share my repertoire or much of my knowledge of the game:
if I'm going to have any shot of winning anything big, surprise value will
do a lot to offset the age problem.
Talent, of course, is a separate issue.
|
| |
Date: 23 Nov 2005 08:29:34
From: JEH
Subject: Re: Hydra crushes Pono in 29 moves
|
> If Hydra can crush them like this, so can a human. We just need to figure
> out how Hydra "thinks."
Computers don't think. Thinking is a flawed process that Humans do
"Kill all the Humans, kill all the humans ..." - Bender
|
| |
Date: 23 Nov 2005 10:05:23
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon <[email protected] > wrote:
> In a Tarrasch French!!
>
> The machine just whacked the poor slob GM.
>
> No endgame required!
Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
open file) looked deeply, deeply dubious. I wonder what he thought the
advantages were because, to a patzer like me, it looks like he spent three
moves putting his king in the most dangerous part of the board. He must
have felt there was some compensation for this somewhere.
Dave.
--
David Richerby Poetic Atom Bomb (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ weapon of mass destruction but it's
in verse!
|
| | |
Date: 23 Nov 2005 10:38:17
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> The machine just whacked the poor slob GM.
>>
>> No endgame required!
>
> Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> open file) looked deeply, deeply dubious. I wonder what he thought the
> advantages were because, to a patzer like me, it looks like he spent three
> moves putting his king in the most dangerous part of the board. He must
> have felt there was some compensation for this somewhere.
When I see GMs doing that, I refuse to quit training at chess for at least a
little while.
If Hydra can crush them like this, so can a human. We just need to figure
out how Hydra "thinks."
|
| | | |
Date: 23 Nov 2005 20:08:04
From: =?ISO-8859-1?Q?Bj=F6rn_Holzhauer?=
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
> When I see GMs doing that, I refuse to quit training at chess for at least a
> little while.
>
> If Hydra can crush them like this, so can a human. We just need to figure
> out how Hydra "thinks."
>
We basically know how Hydra "thinks", it just considers all options and
evaluates the resulting positions. And Hydra does this *really* quickly.
Oh, and it will always (unless there's some software/hardware problem,
of course) follow this algorithm perfectly, it will never overlook a
legal move and it will never evaluate a position other than foreseen by
the position-evaluation-algorithm.
Unless you can evaluate 1000s of positions a second, calculate all
possible continuation in a position to 10 halfmoves depth in a couple of
seconds and can do so without overlooking any legal moves (even 10
halfmoves down the road in all the 100,000s of variations) that is
rather difficult to immitate.
Just to ask, how do you do your encryption when sending encrypted
information via the internet (e.g. when sending credit card details)? Do
you do the calculations in your head based on the public key of the
website? If you think that that's at least a feasible option for you and
wouldn't take you more than a second or two, then go on with thinking
that you could play like a computer.
Bj�rn
|
| | | | |
Date: 24 Nov 2005 15:23:13
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
=?ISO-8859-1?Q?Bj=F6rn_Holzhauer?= <[email protected] > wrote:
> We basically know how Hydra "thinks", it just considers all options and
> evaluates the resulting positions. And Hydra does this *really* quickly.
> Oh, and it will always (unless there's some software/hardware problem,
> of course) follow this algorithm perfectly, it will never overlook a
> legal move and it will never evaluate a position other than foreseen by
> the position-evaluation-algorithm.
Actually, I'd argue that it will never follow the algorithm perfectly
because there will always be bugs. But, by now, most of those bugs that
remain are likely to be fairly insignificant.
Dave.
--
David Richerby Portable Fluorescent Tool (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a screwdriver but it'll hurt your
eyes and you can take it anywhere!
|
| | | | |
Date: 23 Nov 2005 20:39:23
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> When I see GMs doing that, I refuse to quit training at chess for at
>> least a little while.
>>
>> If Hydra can crush them like this, so can a human. We just need to
>> figure out how Hydra "thinks."
>>
>
> We basically know how Hydra "thinks", it just considers all options and
> evaluates the resulting positions. And Hydra does this *really* quickly.
> Oh, and it will always (unless there's some software/hardware problem, of
> course) follow this algorithm perfectly, it will never overlook a legal
> move and it will never evaluate a position other than foreseen by the
> position-evaluation-algorithm.
Right there we have some good advice: "never overlook a legal move."
Seriously, the moves made by Hydra can be reverse-engineered to see what
factors we may be overlooking.
> Unless you can evaluate 1000s of positions a second, calculate all
> possible continuation in a position to 10 halfmoves depth in a couple of
> seconds and can do so without overlooking any legal moves (even 10
> halfmoves down the road in all the 100,000s of variations) that is rather
> difficult to immitate.
I'm looking more at examining the moves themselves and exploring how a human
might arrive at them using our methods.
> Just to ask, how do you do your encryption when sending encrypted
> information via the internet (e.g. when sending credit card details)? Do
> you do the calculations in your head based on the public key of the
> website? If you think that that's at least a feasible option for you and
> wouldn't take you more than a second or two, then go on with thinking that
> you could play like a computer.
I can learn from the moves the computer plays. I also said that I could
*try* to play like the computer.
Like the old joke where two guys are being chased by a bear in the forest,
and one asks the other why he's running since the bear will catch him
anyway. He says "I don't have to outrun the bear; I just have to outrun
you."
|
| | | | | |
Date: 24 Nov 2005 00:09:10
From: =?ISO-8859-1?Q?Bj=F6rn_Holzhauer?=
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
>>>When I see GMs doing that, I refuse to quit training at chess for at
>>>least a little while.
>>>
>>>If Hydra can crush them like this, so can a human. We just need to
>>>figure out how Hydra "thinks."
>>>
>>
>>We basically know how Hydra "thinks", it just considers all options and
>>evaluates the resulting positions. And Hydra does this *really* quickly.
>>Oh, and it will always (unless there's some software/hardware problem, of
>>course) follow this algorithm perfectly, it will never overlook a legal
>>move and it will never evaluate a position other than foreseen by the
>>position-evaluation-algorithm.
>
>
> Right there we have some good advice: "never overlook a legal move."
>
> Seriously, the moves made by Hydra can be reverse-engineered to see what
> factors we may be overlooking.
>
That's eays: "We" (as in humans) tend to overlook some move somewhere
down the road. It's a simple matter of calculating faster and more
accurately than we are capable of doing.
What the computer overlooks (even nowadays) positional elements that are
not fully captured by it's algorithms and of course it every now and
then falls into the traps like e.g. evaluating that it's up by X points
instead of asking "I have so and so much material more than my opponent,
but how do I actually win" (although endgame tables and hashtables help
a bit with that, particularly in endgames).
Looking at good moves to see why they are good is a good idea.
Trying to play "like a computer" doesn't work if it's meant to mean that
one would approach chess in the same way a computer does (because humans
simply can't do some of the things a computer can do). One can of course
try to immitate it's style, but it's just silly. One would limit one's
skills in one way without having the ability to make up for it the same
way a computer can (i.e. calculating power).
As a human one simply has to arrive at the same good moves the computer
may find in a human way - i.e. forming plans, selecting only a few
candidate moves that one really analyzes etc. - sometimes this approach
is better (for the time being) than what the computer manages to do,
sometimes the approaches lead you to the same conclusion and sometimes
the computer will arrive at a better conclusion. However there is little
to be done about that.
|
| | | | | | |
Date: 24 Nov 2005 00:50:58
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> Right there we have some good advice: "never overlook a legal move."
>>
>> Seriously, the moves made by Hydra can be reverse-engineered to see what
>> factors we may be overlooking.
>>
>
> That's eays: "We" (as in humans) tend to overlook some move somewhere down
> the road. It's a simple matter of calculating faster and more accurately
> than we are capable of doing.
We do that through pattern recognition and efficiency. For example, I see
computers playing the Scotch instead of the Lopez. This tells me that if
I'm superior to my peers (a requirement for a world champion), I should be
able to do the same and in doing so carry less risk.
I may even prove that the Scotch offers just as many winning chances, led by
the knowledge that Hydra beats the top GMs with it. Then I look for the
motifs in its play in the scotch, etc.
> What the computer overlooks (even nowadays) positional elements that are
> not fully captured by it's algorithms and of course it every now and then
> falls into the traps like e.g. evaluating that it's up by X points instead
> of asking "I have so and so much material more than my opponent, but how
> do I actually win" (although endgame tables and hashtables help a bit with
> that, particularly in endgames).
There are openings like the Poisoned Pawn where you can't trust computer
analysis. That itself is useful, if I find a win in a line that my opponent
may have dismissed because the computer said it was -2.38 when in fact I was
winning. That means my opponent has to think for himself even at home.
> Looking at good moves to see why they are good is a good idea.
>
> Trying to play "like a computer" doesn't work if it's meant to mean that
> one would approach chess in the same way a computer does (because humans
> simply can't do some of the things a computer can do). One can of course
> try to immitate it's style, but it's just silly. One would limit one's
> skills in one way without having the ability to make up for it the same
> way a computer can (i.e. calculating power).
I find that computers tend to win by getting sharp positional advantages
through short-term surprise tactics (like a Ng4 that crushes f2 when it
didn't appear worth preventing). The computer seems to fight one battle
every game, for one key square, and once it wins that battle, the roof caves
in. They are a lot like Nimzovich in that regard.
> As a human one simply has to arrive at the same good moves the computer
> may find in a human way - i.e. forming plans, selecting only a few
> candidate moves that one really analyzes etc. - sometimes this approach is
> better (for the time being) than what the computer manages to do,
> sometimes the approaches lead you to the same conclusion and sometimes the
> computer will arrive at a better conclusion. However there is little to be
> done about that.
My strategy is to book up as much as possible, and computers are very
valuable for doing this, because unlike the past, I can expand my repertoire
more accurately by checking continuations to see if there is anything that
can keep pace with the machines after I leave my book.
That I am playing against opponents who often do the same only accelerates
this process.
|
| | | | | | |
Date: 24 Nov 2005 00:17:55
From: Antonio Torrecillas
Subject: Re: Hydra crushes Pono in 29 moves
|
En/na Bj�rn Holzhauer ha escrit:
> Trying to play "like a computer" doesn't work if it's meant to mean that
> one would approach chess in the same way a computer does (because humans
> simply can't do some of the things a computer can do). One can of course
> try to immitate it's style, but it's just silly.
Bjorn, ... let's try!
First turn off Ray Gordon, wait for some seconds and then turn on him.
If, after that, He writes something interesting related to chess study,
... I would consider He can act as a computer.
AT (just joking)
|
| | | | | | | |
Date: 24 Nov 2005 18:07:18
From: Toni Lassila
Subject: Re: Hydra crushes Pono in 29 moves
|
On Thu, 24 Nov 2005 00:17:55 +0100, Antonio Torrecillas
<[email protected] > wrote:
>First turn off Ray Gordon, wait for some seconds and then turn on him.
I'm sorry, if I wanted to turn Ray Gordon on I'd read his book about
seduction. I have no intention of doing so.
|
| | | |
Date: 23 Nov 2005 06:38:10
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
>
> >> The machine just whacked the poor slob GM.
> >>
> >> No endgame required!
> >
> > Ponoiov's 12... Kd7?! 13... Kc7 18... Kb7? (moving the king onto an
> > open file) looked deeply, deeply dubious. I wonder what he thought the
> > advantages were because, to a patzer like me, it looks like he spent three
> > moves putting his king in the most dangerous part of the board. He must
> > have felt there was some compensation for this somewhere.
>
> When I see GMs doing that, I refuse to quit training at chess for at least a
> little while.
>
> If Hydra can crush them like this, so can a human. We just need to figure
> out how Hydra "thinks."
This is, of course, a highly debatable issue.
Humans and machines _are_ different when it
comes to specific attributes, traits, skills
and so on. Chess, of course, entails quite a
multitude of such relevant dimensions inter-
acting with one another in actual OTB play.
For example, machines are way more effective at
memorizing opening lines (or sequential pat-
terns, if you like) than humans. It is the
equivalent of allowing one of the OTB parti-
cipants to play "open book" while his opponent
is not allowed to do so but must rely on the
memory of a mere...mortal! 8 >) If the machine
uses ending tablebases, the same can be said
about them as well.
Pitting a human against a machine in OTB play
entails matching "entities" that exhibit relative
strengths and weaknesses that do _not_ match
one another per relevant dimension (e.g., ability
to memorize opening lines). However, in a "black
box" sense, one may try to deduce plausible con-
clusions just by analyzing such games post mortem
(especially if the time controls include a "sudden"
death provision 8 >)).
There may be specific aspects of a machine's
"thinking" process that may be interesting or
even useful to the chess OTB player. Nevertheless,
_emulating_ them may very well prove impossible
due to our limitations as humans...
Major Cat
|
| | | | |
Date: 23 Nov 2005 12:04:01
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> When I see GMs doing that, I refuse to quit training at chess for at
>> least a
>> little while.
>>
>> If Hydra can crush them like this, so can a human. We just need to
>> figure
>> out how Hydra "thinks."
>
> This is, of course, a highly debatable issue.
> Humans and machines _are_ different when it
> comes to specific attributes, traits, skills
> and so on. Chess, of course, entails quite a
> multitude of such relevant dimensions inter-
> acting with one another in actual OTB play.
>
> For example, machines are way more effective at
> memorizing opening lines (or sequential pat-
> terns, if you like) than humans.
Yet they only memorize what a human gives them.
> It is the
> equivalent of allowing one of the OTB parti-
> cipants to play "open book" while his opponent
> is not allowed to do so but must rely on the
> memory of a mere...mortal! 8>) If the machine
> uses ending tablebases, the same can be said
> about them as well.
A human player has a "book" as well, however.
> Pitting a human against a machine in OTB play
> entails matching "entities" that exhibit relative
> strengths and weaknesses that do _not_ match
> one another per relevant dimension (e.g., ability
> to memorize opening lines). However, in a "black
> box" sense, one may try to deduce plausible con-
> clusions just by analyzing such games post mortem
> (especially if the time controls include a "sudden"
> death provision 8>)).
>
> There may be specific aspects of a machine's
> "thinking" process that may be interesting or
> even useful to the chess OTB player. Nevertheless,
> _emulating_ them may very well prove impossible
> due to our limitations as humans...
I look at it more as a video game. In video games, if you make a mistake,
you lose the game. Same in chess. As you stop repeating mistakes, you
master the game and get the high score.
Fact is, these machines are playing better moves than we are, and figuring
out how they determine these moves are better is only logical.
|
| | | | | |
Date:
From: Martin Brown
Subject: Re: Hydra crushes Pono in 29 moves
|
|
| | | | | | |
Date: 23 Nov 2005 20:32:10
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>>>>If Hydra can crush them like this, so can a human. We just need to
>>>>figure
>>>>out how Hydra "thinks."
>
> Hydra computes the situation out to an enormous depth. No human can
> emulate that approach in the same way unless the moves are forced.
May the ghost of Mikhail Tal strike you down!
> Humans short circuit this tedious calculation by pattern matching against
> various motifs and long term strategic goals.
So do computers.
>Tactically even the very best grandmaster can be out manouevered by
>computational hardware now.
I've actually beaten some 3000-rated online monsters (top engines with books
similar to mine) with tactics. Once I got one to fall into a mate-in-seven
because it didn't see the danger and therefore cut off its search after 0.5
seconds.
Put the computer in something like the Muzio Gambit and it gets every bit as
lost as a human without its book. It does, however, also show many more
resources for Black that weren't previously considered.
>This is especially true of Hydra which tries to keep the game move
>permutations high and complex. Precisely the situations where humans tend
>to err and raw computational power is optimised.
Yes, Hydra could probably win many games just by sacrificing a piece for one
or two passers and then rely on its trickery to bring home the point.
> GMs Norwood & Harston constructed some puzzles designed to test human vs
> machine abilities with high discrimination in 1992 (see New Scientist
> #1889 4sept 1993). The human detecting ones will still work today.
>
> Humans would still fail on the human hostile positions,
It seems like every position is hostile to humans now.
>but computer chess has improved by a huge amount since then. I am not
>convinced that computers like Deep Blue or Hydra would be troubled by any
>of them now. Shredder saw through one of them instantly - the same example
>of dumb machine behaviour that Penrose uses in his book "Shadows of the
>Mind".
These computers are becoming more and more like spellcheckers for many parts
of the game.
>>>This is, of course, a highly debatable issue.
>>>Humans and machines _are_ different when it
>>>comes to specific attributes, traits, skills
>>>and so on. Chess, of course, entails quite a
>>>multitude of such relevant dimensions inter-
>>>acting with one another in actual OTB play.
>>>
>>>For example, machines are way more effective at
>>>memorizing opening lines (or sequential pat-
>>>terns, if you like) than humans.
>>
>> Yet they only memorize what a human gives them.
>
> Not any more they don't. They learn from experience.
To a point. Some of these machines are online, learning from us. We,
however, in turn, also learn from them.
>>>It is the
>>>equivalent of allowing one of the OTB parti-
>>>cipants to play "open book" while his opponent
>>>is not allowed to do so but must rely on the
>>>memory of a mere...mortal! 8>) If the machine
>>>uses ending tablebases, the same can be said
>>>about them as well.
>>
>> A human player has a "book" as well, however.
>
> But an imperfect memory,
Speak for yourself. When was the last time you forgot how to tie your
shoes?
>and an ever present tiny risk of over looking one out of the many
>permutations of possible move sequences that leads to disaster.
Computers can be misprogrammed as well. They are very vulnerable to
novelties in that regard, which is why most of them don't play just one
line.
>> Fact is, these machines are playing better moves than we are, and
>> figuring out how they determine these moves are better is only logical.
>
> There is an unstated assumption here that the only way to win is to play
> the computer at its own game. That is very unlikely to work now.
Actually I was thinking more of learning from getting crushed by the
computer at its own game.
Why should I pay a top GM $100 an hour for sparring that is inferior to what
I can get from Shredder? If I'm not st enough to learn from the games
without GM instruction, then I'm not going to become a GM anyway.
> The computer thrives on very sharp tactical lines where exact and
> exhaustive calculation many plies deep will always win out.
Tal showed that one does not need to calculate specific variations to
outthink an opponent. Tactical pressure is as real as positional pressure.
Many great tactical players (Alekhine and Morphy as well as Tal, all of whom
were world champions) used to play on faith a great deal.
>Humans need to focus on long term stanglehold positional play to beat or at
>least draw with a computer.
I haven't found that to be the case and I don't find that strategy to be
very successful.
>If you are lucky the computer doesn't see the net closing inside it's look
>ahead until it is already too late.
>
> Unfortunately Kasparov did not make much headway with anti-computer play
> against Deep Blue (but hiw early games showed promise).
I'd rather play perfect chess and put the burden on the machine to match it.
This is also why I believe in the 1/x rule (each move of the game has a
value of 1/x in importance, with x representing the move number). If you
play perfect for 30 moves out of the gate, when the computer does make
mistakes, you can often get a decisive advantage straight out of the
opening, or win "wire to wire."
I can keep most top computers at bay for about 15-20 moves before they crush
me tactically.
|
| | | | | | | |
Date: 24 Nov 2005 15:08:34
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon <[email protected] > wrote:
> May the ghost of Mikhail Tal strike you down!
He tried but his hand just passed right through me.
Dave.
--
David Richerby Permanent Tool (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ hammer but it'll be there for ever!
|
| | | | | | | |
Date:
From: Martin Brown
Subject: Re: Hydra crushes Pono in 29 moves
|
|
| | | | | | | | |
Date: 24 Nov 2005 18:56:53
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
> Fritz would get to ply 10/25 in well under half a second on my machine.
> Although it is just possible that it could miss a mate in 7, but the
> forced moves and checks typical of a mating sequence would usually mean
> that it did not.
This machine cut off its search for some odd reason. Maybe it wasn't mate
at the beginning but the computer just sped it up.
>>>This is especially true of Hydra which tries to keep the game move
>>>permutations high and complex. Precisely the situations where humans tend
>>>to err and raw computational power is optimised.
>>
>> Yes, Hydra could probably win many games just by sacrificing a piece for
>> one or two passers and then rely on its trickery to bring home the point.
>
> More to the point it can afford to keep the move permutations as high as
> possible without ever simplifying the position unnecessarily. In effect
> waiting for the human to make a tiny mistake on its home ground...
Humans can do the same to computers if their understanding is sufficient. A
"perfect" human player would put the burden on the machine to see
everything.
>>>GMs Norwood & Harston constructed some puzzles designed to test human vs
>>>machine abilities with high discrimination in 1992 (see New Scientist
>>>#1889 4sept 1993). The human detecting ones will still work today.
>>>
>>>Humans would still fail on the human hostile positions,
>>
>> It seems like every position is hostile to humans now.
>
> Designer puzzle positions can be constructed where it is extremely hard
> for a human to see the only winning line.
Unfortunately that helps us little in practice.
>>>>Yet they only memorize what a human gives them.
>>>
>>>Not any more they don't. They learn from experience.
>>
>> To a point. Some of these machines are online, learning from us. We,
>> however, in turn, also learn from them.
>
>>>>A human player has a "book" as well, however.
>>>
>>>But an imperfect memory,
>>
>> Speak for yourself. When was the last time you forgot how to tie your
>> shoes?
>
> I think you are a troll. My point here is that a computer can now have a
> few TB of tablebases and opening books available at instant accurate
> recall.
Leko has a book that rivals that of any machine.
Maybe he's a troll too.
>>>and an ever present tiny risk of over looking one out of the many
>>>permutations of possible move sequences that leads to disaster.
>>
>> Computers can be misprogrammed as well. They are very vulnerable to
>> novelties in that regard, which is why most of them don't play just one
>> line.
>
> Programming errors are not unknown. But there is no reason for a computer
> algorithm to be anything other than deterministic whereas a human player
> may see or overlook something randomly at any time. Especially at the end
> of a match and under time pressure.
Computers can overlook things too. They are far from perfect. If they
were, opening theory would be out at move 50 by now.
>>>>Fact is, these machines are playing better moves than we are, and
>>>>figuring out how they determine these moves are better is only logical.
>>>
>>>There is an unstated assumption here that the only way to win is to play
>>>the computer at its own game. That is very unlikely to work now.
>>
>> Actually I was thinking more of learning from getting crushed by the
>> computer at its own game.
>
> You can probably learn a lot faster by being beaten about 3/4 of the time.
> And then examining the games with a full strength analysis later.
I prefer the strongest possible opposition when learning or training. When
practicing, I play against humans.
> Your approach smacks of learning to box by going 12 rounds with Frank
> Bruno on a regular basis.
Where injuries aren't a factor.
>> Why should I pay a top GM $100 an hour for sparring that is inferior to
>> what I can get from Shredder? If I'm not st enough to learn from the
>> games without GM instruction, then I'm not going to become a GM anyway.
>
> That seems highly likely.
Since I am self-taught, I should be able to use the machines to make GM.
Humans tend to avoid repetition of mistakes.
>>>The computer thrives on very sharp tactical lines where exact and
>>>exhaustive calculation many plies deep will always win out.
>>
>> Tal showed that one does not need to calculate specific variations to
>> outthink an opponent. Tactical pressure is as real as positional
>> pressure. Many great tactical players (Alekhine and Morphy as well as
>> Tal, all of whom were world champions) used to play on faith a great
>> deal.
>
> I would prefer the word intuition. Deep human pattern matching again. The
> same pattern matching skill that makes the best humans still orders of
> magnitude better at Go than computer programs.
Go has more possibilities.
>>>Humans need to focus on long term stanglehold positional play to beat or
>>>at least draw with a computer.
>>
>> I haven't found that to be the case and I don't find that strategy to be
>> very successful.
>
> YMMV
Goldmund does this a lot on ICC. His record is more impressive than most
humans, but hardly a plus score.
>>>If you are lucky the computer doesn't see the net closing inside it's
>>>look ahead until it is already too late.
>>>
>>>Unfortunately Kasparov did not make much headway with anti-computer play
>>>against Deep Blue (but hiw early games showed promise).
>>
>> I'd rather play perfect chess and put the burden on the machine to match
>> it. This is also why I believe in the 1/x rule (each move of the game has
>> a value of 1/x in importance, with x representing the move number).
>
> Who told you that? It is complete and utter garbage!
> 1. e4, d4, c4, or Nb3 and more are all perfectly playable for white.
1. Nb3?
What I said is that the first move from any given position is always the
most important one, because it is NEXT.
What you choose for your first move will impact the game more than any other
move, from the starting position.
> One day a computer program may announce that chess is a first player win
> and the only winning move is my guess 1. e4 but given what happened with
> surprises in endgame table bases it could equally well be 1.g3 or 1.b3
I wasn't saying any specific first move was best.
> Even with continued exponential improvements in computing power I do not
> expect this to happen in my lifetime.
>
>> play perfect for 30 moves out of the gate, when the computer does make
>> mistakes, you can often get a decisive advantage straight out of the
>> opening, or win "wire to wire."
>
> The best chance to get the computer make a mistake is to nudge it towards
> having badly positioned pieces.
Or to take it out of book in a position where it can't see clear to the
middlegame.
>You can usually tell at the end when a plan has been successful as even
>today the engine will suddenly start sacrificing material to delay the mate
>by an extra move.
Yep.
>> I can keep most top computers at bay for about 15-20 moves before they
>> crush me tactically.
>
> Gosh. I am sooo impressed. 15 moves is barely out of the book. I aim to
> hold an even keel against Fritz blitz unleashed for at least the first
> 20-25 moves on opening lines that I am learning. And after that things go
> quickly downhill or I lose on time.
I'm talking about very sharp lines, not "anti computer" lines.
|
| | | | | | | | |
Date: 24 Nov 2005 07:30:37
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
tin Brown wrote:
>
> Ray Gordon wrote:
>
> >>>>>If Hydra can crush them like this, so can a human. We just need to
> >>>>>figure
> >>>>>out how Hydra "thinks."
> >>
> >>Hydra computes the situation out to an enormous depth. No human can
> >>emulate that approach in the same way unless the moves are forced.
> >
> > May the ghost of Mikhail Tal strike you down!
>
> I expect he would agree with me. You are out on a limb.
>
> >>Humans short circuit this tedious calculation by pattern matching against
> >>various motifs and long term strategic goals.
> >
> > So do computers.
>
> They use cached evaluations for move transposition to speed up searching
> but todays computer pattern matching of vague partial motifs and
> spotting potential future patterns is puny compared to human abilities.
>
> >>Tactically even the very best grandmaster can be out manouevered by
> >>computational hardware now.
> >
> > I've actually beaten some 3000-rated online monsters (top engines with books
> > similar to mine) with tactics. Once I got one to fall into a mate-in-seven
> > because it didn't see the danger and therefore cut off its search after 0.5
> > seconds.
>
> In your dreams? Which engine? Show us the game!
>
> Fritz would get to ply 10/25 in well under half a second on my machine.
> Although it is just possible that it could miss a mate in 7, but the
> forced moves and checks typical of a mating sequence would usually mean
> that it did not.
>
> >>This is especially true of Hydra which tries to keep the game move
> >>permutations high and complex. Precisely the situations where humans tend
> >>to err and raw computational power is optimised.
> >
> > Yes, Hydra could probably win many games just by sacrificing a piece for one
> > or two passers and then rely on its trickery to bring home the point.
>
> More to the point it can afford to keep the move permutations as high as
> possible without ever simplifying the position unnecessarily. In effect
> waiting for the human to make a tiny mistake on its home ground...
>
> >>GMs Norwood & Harston constructed some puzzles designed to test human vs
> >>machine abilities with high discrimination in 1992 (see New Scientist
> >>#1889 4sept 1993). The human detecting ones will still work today.
> >>
> >>Humans would still fail on the human hostile positions,
> >
> > It seems like every position is hostile to humans now.
>
> Designer puzzle positions can be constructed where it is extremely hard
> for a human to see the only winning line.
Yes, anti-positional solidity can be
viewed as a human player's...cryptonite!
>
> >>>Yet they only memorize what a human gives them.
> >>
> >>Not any more they don't. They learn from experience.
> >
> > To a point. Some of these machines are online, learning from us. We,
> > however, in turn, also learn from them.
>
> >>>A human player has a "book" as well, however.
> >>
> >>But an imperfect memory,
> >
> > Speak for yourself. When was the last time you forgot how to tie your
> > shoes?
>
> I think you are a troll. My point here is that a computer can now have a
> few TB of tablebases and opening books available at instant accurate
> recall.
>
> >>and an ever present tiny risk of over looking one out of the many
> >>permutations of possible move sequences that leads to disaster.
> >
> > Computers can be misprogrammed as well. They are very vulnerable to
> > novelties in that regard, which is why most of them don't play just one
> > line.
>
> Programming errors are not unknown. But there is no reason for a
> computer algorithm to be anything other than deterministic whereas a
> human player may see or overlook something randomly at any time.
> Especially at the end of a match and under time pressure.
>
> Note to chess engine builders here. I would like an option to add a
> small random number to final node evaluations to avoid this determinism.
> Otherwise I find I can play the same game verbatim against the machine.
Curious, would you want to see such a
feature in computer _analysis_ modes as
well?
>
> >>>Fact is, these machines are playing better moves than we are, and
> >>>figuring out how they determine these moves are better is only logical.
> >>
> >>There is an unstated assumption here that the only way to win is to play
> >>the computer at its own game. That is very unlikely to work now.
> >
> > Actually I was thinking more of learning from getting crushed by the
> > computer at its own game.
>
> You can probably learn a lot faster by being beaten about 3/4 of the
> time. And then examining the games with a full strength analysis later.
>
> Your approach smacks of learning to box by going 12 rounds with Frank
> Bruno on a regular basis.
> >
> > Why should I pay a top GM $100 an hour for sparring that is inferior to what
> > I can get from Shredder? If I'm not st enough to learn from the games
> > without GM instruction, then I'm not going to become a GM anyway.
>
> That seems highly likely.
>
> >>The computer thrives on very sharp tactical lines where exact and
> >>exhaustive calculation many plies deep will always win out.
> >
> > Tal showed that one does not need to calculate specific variations to
> > outthink an opponent. Tactical pressure is as real as positional pressure.
> > Many great tactical players (Alekhine and Morphy as well as Tal, all of whom
> > were world champions) used to play on faith a great deal.
>
> I would prefer the word intuition. Deep human pattern matching again.
> The same pattern matching skill that makes the best humans still orders
> of magnitude better at Go than computer programs.
Tal's playing style proved effective against
some _human_ opposition, which is fair and
good. Matching Tal against, say Hydra, may
have been a completely different proposition...
>
> >>Humans need to focus on long term stanglehold positional play to beat or at
> >>least draw with a computer.
> >
> > I haven't found that to be the case and I don't find that strategy to be
> > very successful.
>
> YMMV
>
> >>If you are lucky the computer doesn't see the net closing inside it's look
> >>ahead until it is already too late.
> >>
> >>Unfortunately Kasparov did not make much headway with anti-computer play
> >>against Deep Blue (but hiw early games showed promise).
> >
> > I'd rather play perfect chess and put the burden on the machine to match it.
> > This is also why I believe in the 1/x rule (each move of the game has a
> > value of 1/x in importance, with x representing the move number).
>
> Who told you that? It is complete and utter garbage!
> 1. e4, d4, c4, or Nb3 and more are all perfectly playable for white.
>
> One day a computer program may announce that chess is a first player win
> and the only winning move is my guess 1. e4 but
> given what happened with
> surprises in endgame table bases
This is an intriguing area. How come
us hobbyists do not hear much more
about these developments. I mean, the
conceptual implications could prove
quite...humbling!
>it could equally well be 1.g3 or 1.b3
>
> Even with continued exponential improvements in computing power I do not
> expect this to happen in my lifetime.
>
> > play perfect for 30 moves out of the gate, when the computer does make
> > mistakes, you can often get a decisive advantage straight out of the
> > opening, or win "wire to wire."
>
> The best chance to get the computer make a mistake is to nudge it
> towards having badly positioned pieces. You can usually tell at the end
> when a plan has been successful as even today the engine will suddenly
> start sacrificing material to delay the mate by an extra move.
> >
> > I can keep most top computers at bay for about 15-20 moves before they crush
> > me tactically.
>
> Gosh. I am sooo impressed. 15 moves is barely out of the book. I aim to
> hold an even keel against Fritz blitz unleashed for at least the first
> 20-25 moves on opening lines that I am learning. And after that things
> go quickly downhill or I lose on time.
>
> Regards,
> tin Brown
|
| | | | | | | |
Date: 24 Nov 2005 00:17:54
From: =?ISO-8859-1?Q?Bj=F6rn_Holzhauer?=
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
>>GMs Norwood & Harston constructed some puzzles designed to test human vs
>>machine abilities with high discrimination in 1992 (see New Scientist
>>#1889 4sept 1993). The human detecting ones will still work today.
>>
>>Humans would still fail on the human hostile positions,
>
>
> It seems like every position is hostile to humans now.
I'm yet to see a computer/program that correctly evaluates this position
this fairly simple position or can make up it's mind about which move it
wants to play as white:
1r6/1P2R3/7p/2pp4/6p1/2k3P1/4KP1P/8 w - - 0 43
wKe2,Re7,Pb7,f2,g3,h2/bKc3,Rb8,Pc5,d5,g4,h6
|
| | | | | | | | |
Date: 24 Nov 2005 10:49:43
From: =?ISO-8859-1?Q?Claus-J=FCrgen_Heigl?=
Subject: Re: Hydra crushes Pono in 29 moves
|
Bj=F6rn Holzhauer wrote:
> I'm yet to see a computer/program that correctly evaluates this positio=
n=20
> this fairly simple position or can make up it's mind about which move i=
t=20
> wants to play as white:
>=20
> 1r6/1P2R3/7p/2pp4/6p1/2k3P1/4KP1P/8 w - - 0 43
> wKe2,Re7,Pb7,f2,g3,h2/bKc3,Rb8,Pc5,d5,g4,h6
Why, is there a move that saves White?
Claus-Juergen
|
| | | | | | |
Date: 23 Nov 2005 15:19:36
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
tin Brown <
|
| | | | | | | |
Date:
From: Martin Brown
Subject: Re: Hydra crushes Pono in 29 moves
|
|
| | | | | |
Date: 23 Nov 2005 09:27:35
From: Major Cat
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
>
> >> When I see GMs doing that, I refuse to quit training at chess for at
> >> least a
> >> little while.
> >>
> >> If Hydra can crush them like this, so can a human. We just need to
> >> figure
> >> out how Hydra "thinks."
> >
> > This is, of course, a highly debatable issue.
> > Humans and machines _are_ different when it
> > comes to specific attributes, traits, skills
> > and so on. Chess, of course, entails quite a
> > multitude of such relevant dimensions inter-
> > acting with one another in actual OTB play.
> >
> > For example, machines are way more effective at
> > memorizing opening lines (or sequential pat-
> > terns, if you like) than humans.
>
> Yet they only memorize what a human gives them.
True. The human programmer provides the
extant, generally acceptable opening
theory (what about refutation lines?)
to the machine. If we visualize a human OTB
player being the recipient of all this
available information about the opening,
well, his memory will certainly be taxed...
Now, if you are thinking of the fact that
generally acceptable opening theory has
been "filtered" by human experts who have
the final say as to what lines get published,
this is a different, although, fascinating
topic.
>
> > It is the
> > equivalent of allowing one of the OTB parti-
> > cipants to play "open book" while his opponent
> > is not allowed to do so but must rely on the
> > memory of a mere...mortal! 8>) If the machine
> > uses ending tablebases, the same can be said
> > about them as well.
>
> A human player has a "book" as well, however.
Yes, of course, he "has" one during OTB play. It
is his very own "closed" book since he cannot consult
published material or notes during the game. If you
are thinking of a duel between asymmetrical "books",
again, this is a different issue of quite some
significance, I should think.
>
> > Pitting a human against a machine in OTB play
> > entails matching "entities" that exhibit relative
> > strengths and weaknesses that do _not_ match
> > one another per relevant dimension (e.g., ability
> > to memorize opening lines). However, in a "black
> > box" sense, one may try to deduce plausible con-
> > clusions just by analyzing such games post mortem
> > (especially if the time controls include a "sudden"
> > death provision 8>)).
> >
> > There may be specific aspects of a machine's
> > "thinking" process that may be interesting or
> > even useful to the chess OTB player. Nevertheless,
> > _emulating_ them may very well prove impossible
> > due to our limitations as humans...
>
> I look at it more as a video game. In video games, if you make a mistake,
> you lose the game. Same in chess. As you stop repeating mistakes, you
> master the game and get the high score.
Depends on how serious the mistake is... 8 >) The
key concept here is, I think, how one evaluates a
move to be a mistake. Machines and humans appear
to be doing this in substantially different ways
which likely conform to their very natures.
>
> Fact is, these machines are playing better moves than we are, and figuring
> out how they determine these moves are better is only logical.
Not necessarily, at least, for now... 8 >) As one
having no particular expertise in how chess programs
work, I have been left with the impression that
machines keep on digging deeper and deeper seeking
killer or, more likely, materially advantageous
tactics!
Major Cat
|
| | | | | | |
Date: 23 Nov 2005 20:35:02
From: Ray Gordon
Subject: Re: Hydra crushes Pono in 29 moves
|
>> I look at it more as a video game. In video games, if you make a
>> mistake,
>> you lose the game. Same in chess. As you stop repeating mistakes, you
>> master the game and get the high score.
>
> Depends on how serious the mistake is... 8>) The
> key concept here is, I think, how one evaluates a
> move to be a mistake. Machines and humans appear
> to be doing this in substantially different ways
> which likely conform to their very natures.
>
>>
>> Fact is, these machines are playing better moves than we are, and
>> figuring
>> out how they determine these moves are better is only logical.
>
> Not necessarily, at least, for now... 8>) As one
> having no particular expertise in how chess programs
> work, I have been left with the impression that
> machines keep on digging deeper and deeper seeking
> killer or, more likely, materially advantageous
> tactics!
I find that computers' "tactical" play has "positional" consequences.
For example, just by playing against it, I've learned how to handle the
sacrifice threat on h7 (sometimes it is sound sometimes not), as well as
knowing when not to allow Ng4 against my kingside.
Sometimes the computer will abandon all sound opening theory to throw three
pieces at the one fatal weak spot in your opening formation. That type of
information is exactly what a player like me needs to sharpen his book.
|
| | | |
Date: 23 Nov 2005 10:45:45
From: Tony Mountifield
Subject: Re: Hydra crushes Pono in 29 moves
|
In article <[email protected] >,
Ray Gordon <[email protected] > wrote:
>
> If Hydra can crush them like this, so can a human. We just need to figure
> out how Hydra "thinks."
Lots, and very, very quickly!
Cheers
Tony
--
Tony Mountifield
Work: [email protected] - http://www.softins.co.uk
Play: [email protected] - http://tony.mountifield.org
|
| |
Date: 23 Nov 2005 07:22:22
From: Eric Hallsworth
Subject: Re: Hydra crushes Pono in 29 moves
|
In article <[email protected] >, Ray Gordon
<[email protected] > writes
>In a Tarrasch French!!
>
>The machine just whacked the poor slob GM.
>
>No endgame required!
>
>
>
And the 'poor slob GM' whacked Fritz9 in round 2!
--
With best wishes from Eric Hallsworth
45 Stretham Road, Wilburton, Cambs CB6 3RX, England
Publisher of Selective Search, the UK's only Computer Chess Magazine, est. 1985.
Computer Chess web pages+photos... http://www.elhchess.demon.co.uk
|
| | |
Date: 23 Nov 2005 10:03:19
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Eric Hallsworth <[email protected] > wrote:
>Ray Gordon <[email protected]> writes
>> In a Tarrasch French!!
>>
>> The machine just whacked the poor slob GM.
>
> And the 'poor slob GM' whacked Fritz9 in round 2!
Huh? The game Ray is referring to is Hydra-Ponoiov in round 2.
Kasimdzhanov played Fritz in round 2 and was only able to draw. This is
the only one of the six games so far that wasn't won by the computer.
Dave.
--
David Richerby Evil Puzzle (TM): it's like an
www.chiark.greenend.org.uk/~davidr/ intriguing conundrum but it's
genuinely evil!
|
| | | |
Date: 23 Nov 2005 11:16:37
From: Luigi Caselli
Subject: Re: Hydra crushes Pono in 29 moves
|
"David Richerby" <[email protected] > ha scritto nel messaggio
news:lJC*[email protected]...
> Eric Hallsworth <[email protected]> wrote:
> >Ray Gordon <[email protected]> writes
> >> In a Tarrasch French!!
> >>
> >> The machine just whacked the poor slob GM.
> >
> > And the 'poor slob GM' whacked Fritz9 in round 2!
>
> Huh? The game Ray is referring to is Hydra-Ponoiov in round 2.
> Kasimdzhanov played Fritz in round 2 and was only able to draw. This is
> the only one of the six games so far that wasn't won by the computer.
There are some little errors...
1. Hydra-Ponoiov was in round 3.
2. Kasimdzhanov played Fritz in round 3
Until now after 9 games:
5 wins for computers
3 draws
1 win for humans
Today last round.
Luigi Caselli
|
| | | | |
Date: 23 Nov 2005 11:39:31
From: David Richerby
Subject: Re: Hydra crushes Pono in 29 moves
|
Luigi Caselli <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> Huh? The game Ray is referring to is Hydra-Ponoiov in round 2.
>> Kasimdzhanov played Fritz in round 2 and was only able to draw. This is
>> the only one of the six games so far that wasn't won by the computer.
>
> There are some little errors...
>
> 1. Hydra-Ponoiov was in round 3.
> 2. Kasimdzhanov played Fritz in round 3
Yes, you're absolutely right -- I must have misread the Chessbase
site. Apologies for the confusion.
Dave.
--
David Richerby Natural Windows (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ graphical user interface but it's
completely natural!
|
| |
Date: 22 Nov 2005 14:31:44
From: [email protected]
Subject: Re: Hydra crushes Pono in 29 moves
|
Ray Gordon wrote:
> In a Tarrasch French!!
>
> The machine just whacked the poor slob GM.
>
> No endgame required!
Please select your reply from the following options:
A) When and where did this happen?
B) Please post the game, or a link to it.
C) I guess that proves that the Tarrasch French is a good opening
for... which side was the computer playing?
D) This must be Ray Gordon's attempt to prove that opening study is
more important than endgame study.
E) Please post a link to pictures proving that the GM is a slob.
:)
(Smiley face included as an aid to the humor impaired)
--Richard
|
| | |
Date: 23 Nov 2005 11:06:58
From: Luigi Caselli
Subject: Re: Hydra crushes Pono in 29 moves
|
<[email protected] > ha scritto nel messaggio
news:[email protected]...
> Ray Gordon wrote:
> > In a Tarrasch French!!
> >
> > The machine just whacked the poor slob GM.
> >
> > No endgame required!
>
> Please select your reply from the following options:
I'm a weak FM from Italy but I try to answer
> A) When and where did this happen?
http://www.ajedrezbilbao.com/FestivalEN.htm
> B) Please post the game, or a link to it.
[Event "2nd Man Machine"]
[Site "Bilbao ESP"]
[Date "2005.11.22"]
[Round "3"]
[White "HYDRA"]
[Black "Ponoiov,R"]
[Result "1-0"]
[BlackElo "2704"]
[EventDate "2005.11.20"]
[ECO "C06"]
1. d4 e6 2. e4 d5 3. Nd2 Nf6 4. e5 Nfd7 5. Bd3 c5 6. c3 Nc6 7. Ne2 a5 8.
O-O c4 9. Bc2 b5 10. b3 Nb6 11. bxc4 bxc4 12. Rb1 Kd7 13. Nf4 Kc7 14. Qf3
Rb8 15. Nh5 Qe7 16. Re1 Ba6 17. Qg3 g6 18. Nf6 Kb7 19. Nde4 Ka7 20. Nd6 Nc8
21. Ba3 Rxb1 22. Rxb1 Nxd6 23. Bc5+ Ka8 24. Ba4 Bb7 25. exd6 Qxf6 26. Bxc6
Bxc6 27. d7 Bxc5 28. Qc7 e5 29. dxc5 1-0
> C) I guess that proves that the Tarrasch French is a good opening
> for... which side was the computer playing?
It's good as many black openings, computer is playing white.
> D) This must be Ray Gordon's attempt to prove that opening study is
> more important than endgame study.
Also Anand says it's true
> E) Please post a link to pictures proving that the GM is a slob.
Sorry for my poor english, but I don't know what slob means...
Luigi Caselli
|
|
