Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Human vs Computer - The End?
1 2 Previous Next  
- - By Linus (***) [at] Date 2007-07-08 14:01
Over the last few days I have been following various comments in the Ehlvest vs Rybka threads, discussing how the handicap could be modified to give the human GM a better chance.

We have seen the current world champion Kramnik loose against Deep Fritz, by far not the best program available today. And we have seen two matches of GM Ehlvest against Rybka, definately the best prorgam available today, with various handicaps like pawn odds, time handicaps, opening book handicaps, end game tablebases handicap, hash table handicap. Now people in other threads are discussing running Rybka on on one processor, on old hardware (like a PII), with smaller hash tables or with a ply limit. The question I have to ask is:

What's the point?

What sense does it make to continue this farce having humans play against computers. We know that no human can win against a good PC running Rybka today. What sense does it make to find a way to handicap the computer so the human has a chance? Give it a rest! This bird is dead! It's over! Move on to something more interesting, like a match Rybka against the best Chessbase product. Or maybe even a match Rybka vs Hydra. I know that it is not easy to set up such matches. At least Vas is trying. But for crying out loud, end this Ehlvest vs Rybka farce now.
Parent - - By Uri Blass (*****) [il] Date 2007-07-08 14:12 Edited 2007-07-08 14:15
We do not know that no human can win against a good PC running Rybka today.

We only know that
Kramnik failed to win against Fritz and that Ehlvest failed against Rybka.

It does not prove that another human cannot win.

Edit:I disagree that comp-comp matches show events are something more interesting

We have comp-comp games all the time(CCRL CEGT SSDF).
We do not have many comp-human games.

Uri
Parent - - By Linus (***) [at] Date 2007-07-08 14:28
I disagree. Kramnik vs DF was not a fair match, Kramnik had a clear advantage the way they played it, and he still lost. Ehlvest had big advantages with all those handicaps and still he lost. I believe that no human (incl. the top GMs) have a chance against an unleashed computer when playing without any handicaps.

Don' get me wrong, I don't want to belittle Ehlvest's performance. Under the given circumstances he is doing a terrific job. The human's time is just up.
Parent - By Michael Waesch [de] Date 2007-07-08 14:40
What a funny discussion. If you followed all the annual human-machine events of Madrid you would know by now that the aera of man is over for years and years, long before anyone even squandered one thought about letting Fritz play against Kramnik. People always seem to neglect that computer processing speed, parallelization of processes - in short: technology can advance almost endlessly putting more and more calculation power to chess engines while man will just be stuck with his own slow brain... At one point in time Chess will be completely solved and gets mostly obsolete and then we head on for newer games and will ask us the same questions about these games too, troughout years, decades and millenia for all time being not learning a single bit.

What a big joke: First man creates a machine that can calculate x-times faster than he can, puts some sort of algorithms into it for having it play Chess and then he refuses for ages to accept that he got easily defeated by that monster. And that´s not only valid for Chess, but for all other areas of technology too. And so, on another historical day those silicon based homuncili will raise against their creators and announce that they don´t need man any longer ...

Mike
Parent - - By Uri Blass (*****) [il] Date 2007-07-08 16:37
It is possible that there are humans who are 600 elo better than kramnik or Ehlvest in playing against computers so results of them in favourable conditions prove nothing.

Uri
Parent - By Hetman (*****) Date 2007-07-08 17:45
if there would have been such a humans they will have to play chess against human on the GM level.
there are no one at the moment. it is possible but only in theory, practicly not.

Hetman
Parent - By Juergen Faas (**) [de] Date 2007-07-08 19:37

>> It is possible that there are humans who are 600 elo better than kramnik or Ehlvest in playing against computers <<


Wanna bet? ;o)
Parent - - By Hetman (*****) Date 2007-07-08 15:04
...either... we do not know that no human can win the race vs the car :-)
Regards
Hetman
Parent - By bluemax (**) [ie] Date 2007-07-08 16:15
I'm not a computer expert or anything, but have been a computer chess fan for over twenty years. Now, for me the last 10-12 years have been the most exciting with better computers and regular increases in chess program strength.

The main point I would like to make is: why not set a standard for the comp-human challenge and stick with it until either it proves itself stronger or weaker then the human players (it plays against)? why keep changing when there is such a disagreement about the results (eg Kramnik-Fritz).

If some one like Vas made a standard (eg dual core, 1GB memory, 5 move opening book, no tablebases etc) and then present his results based on this, then playing with very powerful or mixed (like the WCCC where the system diversity takes away the meaning of the event) would be pointless and competitors would have to follow suit.
Parent - - By Uri Blass (*****) [il] Date 2007-07-08 16:33
There is a difference.

We have times of human and times of cars so we can know that human cannot run faster than cars.
When we talk about computers it is possible that with the right strategy humans can win.

Rating of humans against other humans is meaningless here because I remember that some years ago 2200 player could draw 3 computers in the israeli league when GM's performed worse.

Uri
Parent - By Hetman (*****) Date 2007-07-08 17:31
It is very small difference, none practicly.

Human can win the race vs car in the mountains where is not special road for cars or when on the way is a river and there is no bridge. We can call it special strategy when human will destroy the bridge not to let the car pass. Similar special strategy you proposed for the chess game.

Human can win only in unusual circumstances vs car or chessprogramm.

Rgds
Hetman
Parent - - By Peter Grayson (***) [gb] Date 2007-07-08 17:43
I agree with the post.

All humans can go faster than any motor car. However, put a human in the car and start the engine then it is a different matter, but isn't it a case of the human goes faster in the car than by foot? The car on its own can do nothing useful.

Similarly then with computers and chess for without human input the computer is useless. A human needs to write the code for the computer to use. The human has to switch on the computer and load the program. The code is an extension of human thought so the chess playing computer still only has the ability to play as well as the human input permits.

A faster computer is like a faster car. Fast cars haven't stopped humans racing each other and neither will fast chess playing programs and computers stop humans playing chess against each other either through head to head or computer versus computer media.

Computer only chess is a development of chess by humans just as Grand Prix racing is an extension of the motor car by humans. In each case they can run side by side with human versus human events but the big mistake is to mix them!

PeterG
Parent - - By Peter Grayson (***) [gb] Date 2007-07-08 17:52
As a footnote, I'd like to know which motor car could beat a human at hurdles! :-)

PeterG
Parent - By Fulcrum2000 (****) [nl] Date 2007-07-08 17:56
A tank will do (it would knock down every single one of them, but there are no rules against that on hurdles)
Parent - - By Sesse (****) [no] Date 2007-07-08 18:28
Interestingly enough, as a parallel, there is an annual humans-versus-horses race somewhere in the UK (I forget where). Turns out if you make it a bit longer, and the terrain a bit less flat, the human has a very good chance, and this year, a human even beat all the horses. :-)

/* Steinar */
Parent - By Peter Grayson (***) [gb] Date 2007-07-08 20:03
You're right it did take place but I don't think a horse qualifies as a machine in the man versus machine scenario,
http://news.bbc.co.uk/2/hi/uk_news/wales/mid_/3801177.stm

PeterG
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-07-08 18:06
Ok, but then in this case cancel Wimbledon too. As well as the upcoming NFL season - although the AFC schedule can still be played :)

Vas
Parent - - By Banned for Life (Gold) Date 2007-07-08 18:32
You're right. The NFC schedule can be canceled at any time in favor of just sending Philly to the Super Bowl. :-)

Alan
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-07-10 13:14
Seahawks, Eagles .. this is going in a bad direction. Probably will end with Dallas.

Vas
Parent - - By Banned for Life (Gold) Date 2007-07-10 20:12
All kidding aside, the conventional wisdom, as manifested in numerous local sports books around town, is that the Redskins will come end up third in the NFC East, behind Philly and Dallas. These early predictions are frequently unreliable, but the Redskins winning the division would be a real shocker.

Alan
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-07-12 07:24
Actually, I am quite optimistic this year. The Redskins made some small but important tweaks, the kind that are often overlooked. They addressed their main holes - defensive backfield, defensive backfield, and defensive backfield.

The big question of course is the QB - it's the same exact question for half the teams in the league.

Vas
Parent - - By lkaufman (*****) Date 2007-07-08 18:14
We have proven (at least to my satisfaction) that no human can WIN against Rybka in a match without some handicap, but we have not yet proven that at least some very strong grandmasters might not be able to draw more often than not if that is their goal. In other words, is the superiority of the computer enough to overcome the natural drawish tendency of chess? This is one of the big unanswered questions, and at least one reason for continued curiousity about the human vs. computer events. So I am at least in partial agreement with Uri. Of course the answer to this question depends on colors and time limits, it's not a yes or no thing.
Parent - By Linus (***) [at] Date 2007-07-08 18:21
OK, I see your point. But let the humans prove it without handicaps for the computers. IMHO the ongoing discussion about how to handicap the computer so the GM has a better chance to win or draw is pointless. If there is a human who can draw against an unleashed Rybka more than 50% I would like to see this. I doubt there is one.
Parent - - By Harvey Williamson (*****) Date 2007-07-08 18:29
I think Palm Hiarcs9 may have already proved the end is here a couple of years ago.

Against 2600+ GMs we have played 12 games and lost 0. I think the overall score is 9-3.

http://www.hiarcs.com/palm_hiarcs.htm#phvgm

and also won this tournament:

http://www.hiarcs.com/Games/gmpara.htm
Parent - - By lkaufman (*****) Date 2007-07-08 18:36
Please don't quote blitz/rapid results without identifying them as such. Everyone knows that GMs have been unable to compete with computers for years at these speeds. We're only talking about standard chess now.
Parent - - By Harvey Williamson (*****) Date 2007-07-08 18:39 Edited 2007-07-08 18:43
Who is we?

Is 30+10 on a palm not significant? or 25+0.

On a palm that is probably equivalent to giving Rybka about 1 minute for the games it just played v Ehlvest on your quad.
Parent - - By lkaufman (*****) Date 2007-07-08 18:55
I'm talking about the time for the human player, not the time for the computer. Humans just play much better chess with two hours than with thirty minutes or less. I don't mean to disparage the achievement of Hiarcs Palm, it's just that those games are not relevant to the question of whether humans have been surpassed by computers in serious chess.
Parent - - By Harvey Williamson (*****) Date 2007-07-08 19:06
I guess you could try giving Rybka 1+1 on your quad and the human 30+10 ;-)
Parent - - By lkaufman (*****) Date 2007-07-09 01:10
Actually, now that I think about it, my old programs Rexchess and Socrates (with Don Dailey) had excellent results against Grandmasters, including even a World Championship Candidate, in both blitz and rapid tourneys around 1990 on hardware like a 486 processor at 25 MHz. How would this hardware compare to that of the Palm Hiarcs used? Since this hardware is hundreds of times slower than my quad, I guess it follows that Rybka on my quad could indeed crush grandmasters with 1'+1" vs. 30'+10". I'm only an IM, but I have no chance at all if I play Rybka on my Quad 10' vs. 1'.
Parent - - By Harvey Williamson (*****) Date 2007-07-09 14:32 Edited 2007-07-09 14:36
I think the fastest palms available when these game were played was about 400mhz. If these arm processors at 400mhz equates to say a pentium at 400mhz I am not sure. I suspect the matches were played on fairly fast palms but I don't have those details - so yes looks like faster than a 486.

I found this interesting statistic:

'1996 First microcomputer chess program in history to defeat a FIDE International Master in match play when HIARCS defeated IM Deen Hergott 4-2 in Ottawa, Canada.' This is at classical time controls.

A lot has happened in computer chess in the 10 years since this victory!
Parent - - By Milton (***) [us] Date 2007-07-09 15:40
I remember that match.  It was arranged by a guy named Alan Tomalty who used to post quite often in rec.games.chess.computer and later in CCC under the handle Komputer Korner.  He also used to write reviews about various chess programs in different publications.  He was very impressed with Hiarcs at the time and arranged the contest with Hergott to prove his contention that it was strong enough to beat an IM at normal time controls.  It was a very interesting contest.

Milton
Parent - By Harvey Williamson (*****) Date 2007-07-09 15:53
Hi,

Thanks for your memories of this event. I am not sure how to compare this result to the Ehlvest v Rybka match. Ehlvest is probably 200 elo stronger than Hergott was when the match was played. But I suspect Rybka on Larry's hardware is more than 200 elo stronger than Hiarcs was in 1996 even with half the time.

Interesting stuff!

Best Wishes,

Harvey
Parent - - By Hetman (*****) Date 2007-07-08 19:06
I am not sure if it has been proven, when was the match without handicap ?;-)
You have written in the other thread that the openning book was prepared against Ehlvest no to let him to play the positions he played thousends times. From the openings of the match looks that Ehlvest has not prepared himself against Rybka or computer programm, we have not seen so called anticomputer strategy. That circumstances allow hardly to make the assumption. If both sides were not prepared themselves or both were prepared  for meeting that assumption would be more probable.
Rgds
Hetman
Parent - - By lkaufman (*****) Date 2007-07-08 19:56
I am sure there was a better anti-computer strategy to play if the goal was to make draws, but I am not so sure that there was a better way to play for wins than the way Ehlvest did (given his abilities and experience). I don't know how he could have prepared for this match, he did not anticipate any openings played except the one in the final game, and our very next move, 4...e6, surprised him.
Parent - - By Hetman (*****) Date 2007-07-09 16:43
...4e6. was it in the book ?
Rgds
Hetman
Parent - By lkaufman (*****) Date 2007-07-09 18:07
No, it was three move book. I'm not sure why it surprised him, but he doesn't have a quad, and maybe he prepared before 2.3.2 came out.
Parent - - By h1a8 (***) [us] Date 2007-07-08 20:35
You know what I would like to see? As I am sure others will too.
A super expert (preferrably a GM) at anticomputer play to play against Rybka (with maybe a limited opening book).
I want to see rybka fight against anticomputer play (stonewalls and such).
For I am convinced that no human can win against her playing straight chess.
Sorry Uri, but it has been silly proven that no human can win against top programs in a match using straight chess strategy (non-locked chess game strategy).
The silly proof relies on the two facts: 1. computers can see deeper tactics than humans
                                                    2. and they are decent enough in positional play.
In a 40 move or more game, there is at least a 90% silly probability, that when the human is using straight chess strategy, the computer would see, at least
twice, a tactical combination deeper than the human.

Now I know this isn't a real proof. That is even why I said "silly proof". It might even be nonsense to you. But I am sure most would agree with me. But no real proof doesn't mean that a statement is false. But if someone can actually beat Rybka in a match using anticomputer strategy then by all means they should come forth and challenge Rybka with it. Otherwise they are greatly hindering the progress of computer chess growth. For I am sure that through their games chess engine authors would mainly try to
fix this. Most don't even look at this. Because they don't see it as a real problem (if it exists of course).

So for any GM who is reading this and is thinking about challenging rybka. Please don't not appeal to the fans by playing open straight chess.
Try like hell, using anticomputer strategies, to either draw or win the match. Be greedy and get the money. Trust me, this is what the majority (if not all) of us chess engine fans really want to see.
Parent - - By Michael Waesch [de] Date 2007-07-08 20:52
I disagree. We had already one game in the Ehlvest match that had a strong tendency to stonewall/Fortress position and it was a dead boring draw. People don´t want to see such games, just ask Pablo Restrepo about the comments to his stonewall games. No, people want to see a bloodshed, they want to feel the thrill, they want to see exiting attacking chess and not some wood pushing for desperately keeping a game closed or even stonewalled.

People want Nakamura, not Ehlvest. And I can´t imagine that here a a lot readers who will shout "HURRAY!" if they read another announcement like "Yippe! It´s Ehlvest again!" ...

Mike
Parent - - By Uri Blass (*****) [il] Date 2007-07-08 21:46
I disagree.

You cannot speak for other people
I want humans to do the best that they can and if doing their best is playing in pablo's style then they should do it.

I do not want to see attacking chess unless attacking chess gives practical chances to win the match.

Uri
Parent - - By Michael Waesch [de] Date 2007-07-08 22:03
I disagree. You belong to a clear minority here. The statements in the Ehlvest threads were non-ambiguous. But I will tell Pablo that he can count on you as a supporting and ever watching spectator.

Mike
Parent - By turbojuice1122 (Gold) [us] Date 2007-07-08 22:34
Actually, I would kind of like to see both in separate matches.  I would like to see a match with the attacking chess, i.e. very good attacking chess, with chances for victory for both sides.  Second, I would like to see another match in even conditions (i.e. no time handicap) in which the task of the program is to either try to break through or prevent the human from constructing a stonewall type of defense.  If someone like, say, Kramnik cannot either win a game or draw the match in an 8-game match or longer, then there can be absolutely nobody in the world who can say that humans have much of a chance against top programs in chess.  Obviously the same conclusion arises and has existed for many years now with trying to play "straight-up chess".  However, it would still be nice to see someone like Nakamura play a long match, where he might win one game in 24 and lose all of the others, with maybe a few draws, but the draws and win are quite worth it both from the spectator standpoint for the fighting chess and from the programmers' standpoint, where Vas and Larry can gain a bit of knowledge on exactly how to modify the program to improve it significantly.  Wins by a computer program against Rybka are one thing, but they are a much different style of chess than wins by a human.  Somewhat similarly, a computer/program have achieved something significant if they can consistently prevent a human from playing a stonewall type of drawing defense.
Parent - - By h1a8 (***) [us] Date 2007-07-09 01:44
I disagree. As the majority would want to see stonewalls and other anti-computer strategy.
Just about all of us believe that no human can compete with a top program playing normal chess.
Thus why would we want to see a human lose playing this way (We can just watch ourselves).
We would rather see a GM beating a computer with anti-computer strategy. We can even pick up a couple of pointers
from those games in order to teach us to beat the computer ourselves.

Now don't get me wrong. If it were possible to win using normal chess strategy, then by all means, you're right, people would rather see this.
But since this is not possible (most of us believe), then at least we can pick up a few pointers on how to beat these programs by watching
a GM with anticomputer strategy pick them apart. I'm tired of rybka and others beating the dog crap outta me every time. I want to learn how to beat these
monsters (without having to become a GM). I and the most of us will feel super satisfaction if we are able to draw or beat rybka and others
consistently. This would make my life (and not just my day). So we say, we want to see sucessful anti-strategy against the computer a millions times
more than a good attacking fight that ends up with the human losing (albeit through a blunder).
Parent - - By Linus (***) [at] Date 2007-07-09 03:43
Actually the discussion what "people" want to see is one thing. What they do not want to see is another. I for my part do not want to see another human vs computer match with elaborate handicaps to give the human a chance.
Parent - - By Razor (****) [gb] Date 2007-07-09 04:24
I agree - no handicaps - this makes a nonesense of the objectives of the match.

I for one would prefer to see a tournament played using the rules of chess that the GM is familar with; let Rybka use the same set of rules and run Rybka on hardware that the majority of people use today, i.e., not a quad machine.  The outcome of such a match up would be much more meaningful to those interested in this type of event.
Parent - - By lkaufman (*****) Date 2007-07-09 04:54
The 2-1 time handicap was roughly equivalent to playing with equal time but only using two processors, which would simulate the type of computer many people now have. However, it would have made the games take four hours instead of three, making them less exciting for the spectators and more tiring for GM Ehlvest, so I fail to see any advantage to this proposal. The other limitations had nothing to do with "rules", they were restrictions on the computer only, plus letting Ehlvest have White pieces in every game, none of which can possibly be considered unfamiliar to him. Surely you would not argue that the computer should be required to play twenty moves of mainline opening theory, just because it would be more familiar to the GM?
Parent - - By Razor (****) [gb] Date 2007-07-10 04:44
From the various responses I have seen on this subject, it would seem to me that people want to judge where the software they have bought stands.  For some this is where does it stand against another piece of software, for others it is against humans.  Clearly, the more one deviates away from what is considered 'norm', the more that some form of extrapolation will play a part in any explanation of results.  The more extrapolation one has, the more likely your audience will be in dismissing the results.

Whilst the results of such a match up are important, how the results were derived is even more important.  I see almost everyday comments on this website that seem incredibly stupid or unfair against Rybka.  Using an inferior method will just make this worse.

For what it is worth, I believe (generally) that any of the top-10 chess engines would beat the highest rated GM (currently Anand I think) playing on a standard mid-range laptop under normal tournament controls.  Humans tend to make careless mistakes OTB when they get tired; computers on the otherhand either work or they don't work - there is no grey area!  However, if you were to ask me, would any of the top-10 chess engines have a better understanding of the game than Anand or play more 'beautiful' (artistic) chess than Anand then I am less sure.  My perception is that I would say Anand would be better than most of the top-10 engines in most areas of the game; I am not so sure if this would be the case against Rybka 2.3.2a or Hiarcs 11.1.
Parent - By lkaufman (*****) Date 2007-07-10 05:20
I didn't know that chess engines had any "understanding" of chess at all! Of course Anand (and many other great players) are superior to all chess engines in judgement, planning, creativity (whatever that means), and countless other chess-related attributes. But chess results depend primarily on accuracy of analysis, and in this area no human can compete with the engines on today's hardware (especially Rybka!). It's not a question of the human getting "tired"; it's just that computers can calculate thousands (maybe millions) of times faster than they can. Humans have to be very selective about what they look at, and sometimes their criteria for selection turn out to be wrong.
     Actually I can pretty well prove your statement about understanding. When I got heavily involved with the evaluation function of Rybka after version 2.3.1 came out, I substituted my own judgement for the existing evaluation of Rybka in many areas. You can say that in those areas, version 2.3.2 reflects my "understanding" as compared to version 2.3.1 reflecting Rybka's own understanding (which was partly determined by automated means, it wasn't just Vas making his own judgements on everything). The results are pretty clear -- 2.3.2 is substantially stronger (though I'm not taking full credit for this of course). So there you have it; my "understanding" of the aspects of chess I worked on was proven to be superior to Rybka's, and if that is the case, we needn't wonder whether Anand's understanding would be better too.
Parent - By Hetman (*****) Date 2007-07-09 14:54
That is. No handicups.

So no databases and openings book for the programm or the database and openning book for the human  in the match!
If equal then equal.
Regards
hetman
Parent - - By Michael Waesch [de] Date 2007-07-09 07:46
Well, in that case just get Pablos games, watch and enjoy. There is no grandmaster needed to employ such strategies and if you are pleased by extremely long games where just a Rook hovers on the 2nd/7th rank until the game is drawn. Go, get it and watch it in infinite loop!

Mike
Parent - - By h1a8 (***) [us] Date 2007-07-09 16:00
I would rather see longer games instead of the rapid games he plays.
And many of his wins are due to rybka losing on time. And his anti-computer style isn't real anti-computer chess (its a fake copy).
Maybe its the time controls that forces it to be the real stuff?

There is a fine difference between the way Pablo plays and the way Ernest F., M.D. Pecci plays (author of Chess: A Psychiatrist Matches Wits with Fritz ISBN-10: 1929331045 ).
Dr. Pecci in his book employs specialized barrage positions in which he always attack the computer on the kingside by first building up a specialized barrage position in which to attack from for mostly wins and rarely a draw (As he shows in countless games). He has done this against Fritz 6, Shredder 6, and even Fritz 7. You should check this book out. Its just my strength, tactics, and attack ability is too weak (or I'm too lazy to read the whole book) for me to beat Rybka and others. I'm only a 1700elo at best (1600 on average). Maybe I should read and study the entire book? I would just rather see such strategy on a more newer engine (like rybka).

Anyway in one game against X3D Fritz, Kasparov employed a somewhat similar tactic to tie Fritz up and slowly crush it for an easy win. That game was beautiful to me and not an ounce of boring play. It was very embarrassing for Fritz. But kasparov only used this tactic once in the entire match. Which was just to get back even.  The authors of Fritz probably makes deals with the GM playing (getting them to not play anti-computer chess against Fritz). That is why I think Kasparov and Kramnik (and maybe some others) and instead played normal chess. Either it was a deal or they did it for the fans. But nonetheless, I very much believe that most GM could win easily against the computer employing such tactics.

So Pablo's technique is not the same. Especially playing at the time controls he does. Albeit if he plays at much much longer time controls then he probably could win many.
Assuming he employs the techniques in Dr. Pecci's book.
Up Topic Rybka Support & Discussion / Rybka Discussion / Human vs Computer - The End?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill