hgm wrote:Humans are a done deal. The strongest player is not a Human at this very moment, and Humans will even be less competative in the future due to further hardware advances. Only those who are stuck in the past would see beating Humans as a worthy goal...
But that is all beside the point. "We all" is an erroneous statement even if there is one person on the planet that doesn't agree. And from the people I know, those aiming to beat Humans is a vast minority.
You don't seem to be in touch with reality at all. In the other thread I already introduced the term "solipsistic reality" for this, and this is another clear example.
Note that the only target of beating humans that may be interesting is beating humans with material handicap but in this case I think that you need to use psychological factors like what you expect the opponent not to see to perform better.
Beating more humans with queen handicap is a different type of task than writing a program to perform better in games against other programs.
One reason that most programmers are not interested in the first task is because it is easier to test the second task.
Another reason is that programming something like if I do this and the opponent does not see this seems to be harder task and programmers prefer to do easier things.
Uri
I think this discussion is stupid. Which of you two (Uri and HGM) has a program that can beat all humans even a majority of the games, much less 100%? I _KNOW_ the answer. But it takes a dose of reality for some to accept the answer. Humans are _not_ a "done deal". Anti-computer players are still a _big_ problem for computers.
All this shows is a complete lack of experience (on the part of HGM) in playing against lots of humans, say on ICC, using a computer. It is just a ridiculous and inaccurate statement, unless he is transferring the success of Rybka (which does not win every game against humans) to his program, even though they are not close...
"we all only care about how small our engines are, rather than how strong they play"
So you think _I_ am the only person that is developing a chess engine to beat strong humans?
get real...
He did not say that you are the only one but I believe that the majority of programmers do not care about playing against humans because today it is not a problem to beat the humans that you play against them(and the question if you score 70% or 90% is not very interesting).
Sorry, but that's a load of crap. Most programs today can _not_ beat strong humans with regularity. Certainly not 90% of the time.
Most engines can beat more than 99.999% of humans and if we talk about strong GM's then I see no reason for them to be interested in playing serious matches against programs that are significantly weaker than rybka(no sponsor is going to care if a GM can beat an engine that is not close to the top).
Uri
Who cares about 99%? That has been the case for almost 20 years, since less than 1% of humans are masters, and computers have been rated masters since 1980 or so. It is that other 1% that has proven (and still proves) to be a challenge...
I don't know where you/HGM play chess, but you are not playing the opponents I am used to playing against...
hgm wrote:Indeed, opponent modelling is an entirely different field, not touched by most Chess programmers.
But it is completely irrelevant what is done more frequently; _everything_ that is supported should work flawlessly, even if it is as outlandish as playing the Crazyhouse version of shuffle Shatranj on an ICS.
NOte the "everything" and "should work flawlessly" in your statement. Have I been saying _anything_ that would contradict that idea? Of course not. I want _everything_ to work flawlessly. Including draw offers and claims. Playing against humans who can and do occasionally try to make an illegal move or make an invalid draw claim.
So don't write "everything" when you obviously mean only "some things"...
Otherwise we would not be having this argument.
Furthermore note that FIDE rules only have a very limited jurisdiction, and people playing Gothic Chess, Shatranj, Xiangqi or Shogi are not bound by them even now...
bob wrote:Right. And just like the roof, once again my point goes right over your head and is missed...
Indeed. You should be prepaired for such things as long as your head floats somewhere in the clouds. But you are right about one thing: this discussion _is_ stupid. (And who again brought it on this subject? ) If only because it is totally irrelevant for the question if a supported feature should work for those who want to use it.
So no matter how much you continue to argue, things will be made to work correctly. Bugs will not be tolerated.
hgm wrote:Humans are a done deal. The strongest player is not a Human at this very moment, and Humans will even be less competative in the future due to further hardware advances. Only those who are stuck in the past would see beating Humans as a worthy goal...
But that is all beside the point. "We all" is an erroneous statement even if there is one person on the planet that doesn't agree. And from the people I know, those aiming to beat Humans is a vast minority.
You don't seem to be in touch with reality at all. In the other thread I already introduced the term "solipsistic reality" for this, and this is another clear example.
Note that the only target of beating humans that may be interesting is beating humans with material handicap but in this case I think that you need to use psychological factors like what you expect the opponent not to see to perform better.
Beating more humans with queen handicap is a different type of task than writing a program to perform better in games against other programs.
One reason that most programmers are not interested in the first task is because it is easier to test the second task.
Another reason is that programming something like if I do this and the opponent does not see this seems to be harder task and programmers prefer to do easier things.
Uri
I think this discussion is stupid. Which of you two (Uri and HGM) has a program that can beat all humans even a majority of the games, much less 100%? I _KNOW_ the answer. But it takes a dose of reality for some to accept the answer. Humans are _not_ a "done deal". Anti-computer players are still a _big_ problem for computers.
All this shows is a complete lack of experience (on the part of HGM) in playing against lots of humans, say on ICC, using a computer. It is just a ridiculous and inaccurate statement, unless he is transferring the success of Rybka (which does not win every game against humans) to his program, even though they are not close...
I do not claim that most programs can beat all humans in the majority of the games but I think that most programmers are not interested in that problem.
Programmers of programs that are not top programs do not plan to pay money to strong GM's in order to try to win against their program in a match and I believe that in ICC they are not going to play long time control games against strong GM's so the programs can win most of their games(maybe somebody with pablo's style may draw often but humans usually do not like to play like that and this style can only help strong GM to draw and not to win).
I also think that it is easy to solve the pablo problem by making the program weaker against other programs and I am simply not interested in it(I understood in the rybka forum that the version of rybka that played to win with black against benjamin used anti-symmetric evaluation that is slightly counter productive against other programs).
Testing against other programs is also more easy because it is easy to find programs to test against them.
I believe that today there are many free programs that are already at least 2600 fide rating based on long time control games(if they only get the opportunity to prove it but unfortunately Fide does not allow them to play)
"we all only care about how small our engines are, rather than how strong they play"
So you think _I_ am the only person that is developing a chess engine to beat strong humans?
get real...
He did not say that you are the only one but I believe that the majority of programmers do not care about playing against humans because today it is not a problem to beat the humans that you play against them(and the question if you score 70% or 90% is not very interesting).
Sorry, but that's a load of crap. Most programs today can _not_ beat strong humans with regularity. Certainly not 90% of the time.
Most engines can beat more than 99.999% of humans and if we talk about strong GM's then I see no reason for them to be interested in playing serious matches against programs that are significantly weaker than rybka(no sponsor is going to care if a GM can beat an engine that is not close to the top).
Uri
Who cares about 99%? That has been the case for almost 20 years, since less than 1% of humans are masters, and computers have been rated masters since 1980 or so. It is that other 1% that has proven (and still proves) to be a challenge...
I don't know where you/HGM play chess, but you are not playing the opponents I am used to playing against...
I believe that many free programs can regulary beat masters.
grandmasters at long time control may be a problem but
I think that most programmers do not care about that problem
and if they only make their programs better against other programs without caring about humans then the best GM's will be able to do is maybe drawing most of their games but they will be unable to win a match.
I doubt if today the top GM's that can beat Glaurung in a match
even if you use long time control(assuming that they do not know the book of Glaurung before the games).
"we all only care about how small our engines are, rather than how strong they play"
So you think _I_ am the only person that is developing a chess engine to beat strong humans?
get real...
He did not say that you are the only one but I believe that the majority of programmers do not care about playing against humans because today it is not a problem to beat the humans that you play against them(and the question if you score 70% or 90% is not very interesting).
Sorry, but that's a load of crap. Most programs today can _not_ beat strong humans with regularity. Certainly not 90% of the time.
Most engines can beat more than 99.999% of humans and if we talk about strong GM's then I see no reason for them to be interested in playing serious matches against programs that are significantly weaker than rybka(no sponsor is going to care if a GM can beat an engine that is not close to the top).
Uri
Who cares about 99%? That has been the case for almost 20 years, since less than 1% of humans are masters, and computers have been rated masters since 1980 or so. It is that other 1% that has proven (and still proves) to be a challenge...
I don't know where you/HGM play chess, but you are not playing the opponents I am used to playing against...
I believe that many free programs can regulary beat masters.
grandmasters at long time control may be a problem but
I think that most programmers do not care about that problem
and if they only make their programs better against other programs without caring about humans then the best GM's will be able to do is maybe drawing most of their games but they will be unable to win a match.
I doubt if today the top GM's that can beat Glaurung in a match
even if you use long time control(assuming that they do not know the book of Glaurung before the games).
Uri
Whether or not that is true remains to be seen. But I will guarantee you there will be way too many draws in such a match, a result I believe can be improved on. But not by just more depth.
) If only because it is totally irrelevant for the question if a supported feature should work for those who want to use it.