The Evidence against Rybka

[veritas]

@ Chris (DIM) Whittington


CAN YOU READ

or was my simple question to difficult for you to answer




one could suppose that someone with such a surname might be PANTOMIME minded , your ridiculous arguments turn this subject into one , save them for rybkas theatrical forum

[orgfert]

Our contention is that ideas in Fruit which are paralleled in some way in Rybka are NOT "owned by Fruit" in the first place.

But the specific recipe of how the ideas (which you say are not owned by fruit) are combined is owned by fruit. That recipe is fruit's idea.

Absolutely, and that's the key and where the discussion should eventually be. Unfortunately we are still bogged down in disagreement over the particular importance of individual elements and with confusing global use of the term copying vis a vis ideas/code etc

So one could copy someone else's code having convinced himself all ideas contained therein are public domain, munge a few procedures, maybe add some original ideas and call it a new name. This seems to pass your personal smell test. Yet one is still selling a product with significant portions of a copyrighted(lefted) recipe of public domain ideas. Not kosher in very many people's books (and some laws) including the ICGA's.

Correct me if I've not described your argument rightly.

Even if someone thinks it's OK, does that make it also OK to violate a club rule (ICGA) against the practice?

[hyatt]

First, if you want to use a big word, learn to spell it correctly. With two s letters: dissemble is apparently what you want. however, I responded to your nonsense on RF by pointing out that for the king shelter stuff, fruit did it by computation, rybka did it via a table, the code you posted does not exist, and there is a paragraph in Zach's report that explains this, just like there was for the PST data. The paragraph you almost never found...

The code I posted was real, and is present, and is expressed in an atypical way. We can move on to another piece if you want. So far, however, you are batting zero. I'd think you would get tired of swinging and missing so frequently...

The only one "dissembling" is you...

[hyatt]

Minority Report 5, Not plagiarism unless feature is unique

This is a limited definition of plagiarism. In the Rybka/Fruit case, the principal aspect was the overlap in the set (and specification) of features, not any single feature itself. This has been discussed so often, both in the abstract and in the case here, it is almost pointless to reiterate.

I refer to this post, with the quotations Originality requires neither novelty or uniqueness, and The Appellate Court found that the term original should be read to mean "owes its origin" to a particular author, and not that the work was "startling, novel or unusual, or a marked departure from the past." In the Rybka/Fruit case, the specific choice of evaluation features in Fruit 2.1 "owed its origin" to Letouzey, and various Rybka versions were found (by the ICGA) to derive from this origin.

Apologies, I diverted you from the dictionary/wiki definition of "plagiarism" by changing it slightly as a title.

The definition actually says:

Twentieth-century dictionaries define plagiarism as "wrongful appropriation," "close imitation," or "purloining and publication," of another author's "language, thoughts, ideas, or expressions," and the representation of them as one's own original work,[1][2]

cutting this definition down for our purposes of discussion, we have

Twentieth-century dictionaries define plagiarism as "close imitation," of another author's "ideas" AND the representation of them as one's own original work.

Key to this definition is the term another author's "ideas"

If the "idea" was not Fruit's but "another author's" then it is not plagiarism of Fruit. So, if you can't show the "idea" as what I maybe erroneously describe as "unique" when what I mean is "owned by Fruit" only, then the "plagiarism" charge does not hold and "plagiarism" case collapses. Our contention is that ideas in Fruit which are paralleled in some way in Rybka are NOT "owned by Fruit" in the first place.

You could see this "ownership" concept as a kind of FILTRATION process prior to COMPARISON.

another author's "language, thoughts, ideas, or expressions

That says it all. Source code is a semantic expression of an idea. That is what can NOT ber copied. The key is NOT "idea". The key is "the expresssion of that idea', AKA "source code."

There is NO game-playing code copied. If there was you would put it side by side, as asked a thousand times, and show it. But you don't and can't. The Zach document shows parallel use of IDEAS, different weights, different implementations. The ideas overlap is partial. Not all of Fruit is in Rybka, not all Rybka is in Fruit. Rybka contains own ideas and Fruit ideas that Vas said all along he went forwards and backwards taking. There's no copyright violations and the plagiarism charge collapses on the basis that the cross-over ideas are not owned by Fruit in the first place. You can't plagiarise null move or minimax because they are in general usage, likewise you can't plagiarise used or second hand evaluation ideas.

I find it hard to put it together "side by side". However, I did oblige you by putting it together, Fruit first, then Rybka, which is easier to do and easier to read. You ran and hid, promptly, after some silly constant-folding nonsense...

[BB+]

There is NO game-playing code copied.

If this said "There is no game-playing code literally copied", then I would be more prone to agree. Then again, it seems we don't have a literal copy of any Rybka version that competed in any competition, ever. . And on top of that we can't compare Fruit source to Rybka disassembly literally. Horrors! plagiarism and copyright law thus can't apply here!

Also, the declination to merely "game-playing" is not completely clear to me in general. For instance, suppose a 16th century map-maker X makes a map of "Amerika", complete with the made-up barren island of Fugato. Map-maker Y copies (non-literally) the map, including Fugato. There was "NO" useful information copied. But I could only chuckle if Y claimed (w/o any other evidence) that only Fugato was copied, and the rest was his "original" work. Or more legalistically in modern jargon: the copying of "mistakes" of others is quite significant prima facie evidence that copying occurred on the whole, and the evidential burden would typically shift to the accused to demonstrate that any disputed parts are indeed original. So the non-game-playing code from Fruit 2.1 that appears in Rybka 1.0 Beta is indeed relevant in its implications.

Perhaps Zach is also willing to explain the extraordinary factor differences between Fruit and Rybka. [...]

Factors are only one part of the discussion. If you copy X but change the numbers, you still (non-literally) copied X. Part of the derived work (X+number_changes) is specific to you, and part of it to X. Any percentages depend on the case. As part of the context was PSTs, I might return you to the challenge of generating other PST sets with as few changes from Fruit 2.1 code as Rybka 1.0 Beta requires. It is possible that this "minimal change-set for [functional] equivalence" metric will be used with Fabien/FSF, and perhaps EVAL_COMP will additionally be complemented by such. Such a metric particularly addresses what the "origins" of a program likely are.

[Chris Whittington]

There is NO game-playing code copied.

If this said "There is no game-playing code literally copied", then I would be more prone to agree. Then again, it seems we don't have a literal copy of any Rybka version that competed in any competition, ever. . And on top of that we can't compare Fruit source to Rybka disassembly literally. Horrors! plagiarism and copyright law thus can't apply here!

Also, the declination to merely "game-playing" is not completely clear to me in general. For instance, suppose a 16th century map-maker X makes a map of "Amerika", complete with the made-up barren island of Fugato. Map-maker Y copies (non-literally) the map, including Fugato. There was "NO" useful information copied. But I could only chuckle if Y claimed (w/o any other evidence) that only Fugato was copied, and the rest was his "original" work. Or more legalistically in modern jargon: the copying of "mistakes" of others is quite significant prima facie evidence that copying occurred on the whole, and the evidential burden would typically shift to the accused to demonstrate that any disputed parts are indeed original. So the non-game-playing code from Fruit 2.1 that appears in Rybka 1.0 Beta is indeed relevant in its implications.

Perhaps Zach is also willing to explain the extraordinary factor differences between Fruit and Rybka. [...]

Factors are only one part of the discussion. If you copy X but change the numbers, you still (non-literally) copied X. Part of the derived work (X+number_changes) is specific to you, and part of it to X. Any percentages depend on the case. As part of the context was PSTs, I might return you to the challenge of generating other PST sets with as few changes from Fruit 2.1 code as Rybka 1.0 Beta requires. It is possible that this "minimal change-set for [functional] equivalence" metric will be used with Fabien/FSF, and perhaps EVAL_COMP will additionally be complemented by such. Such a metric particularly addresses what the "origins" of a program likely are.

Perhaps if I spell out the critical point in easy to understand terms, we might actually get an answer from you?

Shakespeare writes "to be or not to be"

Some time later and forgetting about timeout issues, person T translates and writes "sein oder nicht sein"

Some time later person U writes "etre ou n'etre pas", pardon my French

Now, we don't really know if U copied/translated T or S or invented it himself, but we certainly can't say here that U plagiarised T because it is S who wrote the original, even if it were the case that U derived his expression from T, it is not T's expression.

The parallel to Fruit/Rybka, is that plagiarism (the verdict levelled by the icga) is not applicable UNLESS FRUIT OWNS THE ALLEGEDLY PLAGIARISED THING.

That's why we don't talk about plagiarism of null move or minimax. Too common, too often used. Evaluation sub components likewise only you missed that in your reports.

[Chris Whittington]

Our contention is that ideas in Fruit which are paralleled in some way in Rybka are NOT "owned by Fruit" in the first place.

But the specific recipe of how the ideas (which you say are not owned by fruit) are combined is owned by fruit. That recipe is fruit's idea.

Absolutely, and that's the key and where the discussion should eventually be. Unfortunately we are still bogged down in disagreement over the particular importance of individual elements and with confusing global use of the term copying vis a vis ideas/code etc

So one could copy someone else's code having convinced himself all ideas contained therein are public domain, munge a few procedures, maybe add some original ideas and call it a new name. This seems to pass your personal smell test. Yet one is still selling a product with significant portions of a copyrighted(lefted) recipe of public domain ideas. Not kosher in very many people's books (and some laws) including the ICGA's.

Correct me if I've not described your argument rightly.

Even if someone thinks it's OK, does that make it also OK to violate a club rule (ICGA) against the practice?

Basically you are trolling because you are inventing something I never said in any way shape or form. Then you expect me to come back and "correct you". Which is a bore for me to correct something I never suggested in the first place.

I suggest you go search the posts in which I state "..... within the law, including copyright law".

[hyatt]

There is NO game-playing code copied.

If this said "There is no game-playing code literally copied", then I would be more prone to agree. Then again, it seems we don't have a literal copy of any Rybka version that competed in any competition, ever. . And on top of that we can't compare Fruit source to Rybka disassembly literally. Horrors! plagiarism and copyright law thus can't apply here!

Also, the declination to merely "game-playing" is not completely clear to me in general. For instance, suppose a 16th century map-maker X makes a map of "Amerika", complete with the made-up barren island of Fugato. Map-maker Y copies (non-literally) the map, including Fugato. There was "NO" useful information copied. But I could only chuckle if Y claimed (w/o any other evidence) that only Fugato was copied, and the rest was his "original" work. Or more legalistically in modern jargon: the copying of "mistakes" of others is quite significant prima facie evidence that copying occurred on the whole, and the evidential burden would typically shift to the accused to demonstrate that any disputed parts are indeed original. So the non-game-playing code from Fruit 2.1 that appears in Rybka 1.0 Beta is indeed relevant in its implications.

Perhaps Zach is also willing to explain the extraordinary factor differences between Fruit and Rybka. [...]

Factors are only one part of the discussion. If you copy X but change the numbers, you still (non-literally) copied X. Part of the derived work (X+number_changes) is specific to you, and part of it to X. Any percentages depend on the case. As part of the context was PSTs, I might return you to the challenge of generating other PST sets with as few changes from Fruit 2.1 code as Rybka 1.0 Beta requires. It is possible that this "minimal change-set for [functional] equivalence" metric will be used with Fabien/FSF, and perhaps EVAL_COMP will additionally be complemented by such. Such a metric particularly addresses what the "origins" of a program likely are.

Perhaps if I spell out the critical point in easy to understand terms, we might actually get an answer from you?

Shakespeare writes "to be or not to be"

Some time later and forgetting about timeout issues, person T translates and writes "sein oder nicht sein"

Some time later person U writes "etre ou n'etre pas", pardon my French

Now, we don't really know if U copied/translated T or S or invented it himself, but we certainly can't say here that U plagiarised T because it is S who wrote the original, even if it were the case that U derived his expression from T, it is not T's expression.

The parallel to Fruit/Rybka, is that plagiarism (the verdict levelled by the icga) is not applicable UNLESS FRUIT OWNS THE ALLEGEDLY PLAGIARISED THING.

That's why we don't talk about plagiarism of null move or minimax. Too common, too often used. Evaluation sub components likewise only you missed that in your reports.

Silly argument. "to be or not to be" == 6 words. "null-move" = 1 or 2 words depending on how you count them. But beyond that, there are MANY implementations of null-move, which in Crafty for example, is 34 lines of C code in one place in search, plus a few more lines below the HashProbe() code, plus a few more lines in HashProbe() itself (avoid null-move trick.) Stockfish adds a verification. Diep (and a few others) do a double null-move search where the second serves as zugzwang detection for the first.

you are free to use null-move in your program... but you are NOT free to copy my null-move implementation and use it in your program. Why you keep trying to slide that argument by incognito is beyond me, as it won't fly.

null-move, minimax, negamax, are all IDEAS. We are talking about copying actual implementations of those ideas, which is where the problem arises...

[hyatt]

Problem is, you are misapplying copyright law as it applies to computer software. Ideas are not copyrightable. Expressions of ideas (source code) IS.

[mjlef]

Wow, this stuff written by Chris is very deceptive. I grow tried of refusting this
stuff, but will do it one more time before I give up!

>>Perhaps Zach is also willing to explain the extraordinary factor differences between Fruit and Rybka.

Fruit:
static const int RookSemiOpenFileOpening = 10;
static const int RookSemiOpenFileEndgame = 10;
static const int RookOpenFileOpening = 20;
static const int RookOpenFileEndgame = 20;

Rybka:
static const int RookSemiOpenFileOpening = 64; // factor 6.4
static const int RookSemiOpenFileEndgame = 256; // factor 25.6 (25.6 / 6.4 -> factor 4 more as in Fruit)
static const int RookOpenFileOpening = 1035; // factor 51.7 (51.7 / 6.4 -> factor 8 more as in Fruit)
static const int RookOpenFileEndgame = 428; // factor 21.4 (21.4 / 6.4 -> factor 3.3 more as in Fruit)
>>

Chris totally ignores the fact that Rybka uses units of 3200 or so for a pawn where Fruit uses 100=pawn.
This means the Rybka values should naturally be 32 times bigger than the Fruit values. So trying to
compare them make the difference much greater than it actually is. Rybka shifts the final value by 5 bits
at the end of the evaluation, which divides by 32, bringing the final values in line. Chris not mentioning
this is deceptive. And anyway, the whole thing is a straw man. The panel did not ever say Rybka was an
exact clone of Fruit. Merely that it was a derivative. A derivative could have changed values, but still
use the evaluation of the same features. The feature evaluation is what was copied, then some values
modified.

If I took Fruit, changed a few values in the eval, and then claimed it as my own, the result would still be
a derivative. And my claim of originaity would be deceptive.

>>Trotsky Date 2011-09-25 09:39
Idea - parallel usage
Implementation - a bit different but being very different in such a simple sub-function is not easy, even
the Hyatt example Crafty code was semantically almost identical
Weights - completely different
Relative balance between the weights - completely different
Chessicly - very different, will lead to different move selections
Copyright - no violations
Plagiarism - concept of open files, semi open files is not owned by Fruit, so plagiarism does not apply
>>

All straw man arguments. Plagarism does still apply when you make a derivative. And Rybka used almost all
the same evaluation fetaures as Fruit. The overlap is astounding and undeniable.

Crafty's evaluation is hugely different from Fruit or Rybka.

Again, if I change a couple of values in Fruit it will choose different moves. But that does not mean the modified
program is not a derivative.

>>
So far we have not studied ONE SINGLE SECTION of the Zach document which itself supports either copyright or plagiarism charges, whether the section was selected by us or selected by Hyatt. PSTs, no good. Pawn shelter, no good. Open and half open files for rook, no good.

You'll find the original on Rybka forum. Stop with the disembling please.
>>

Nonsense. There is huge overlap of all evaluation components.

I have not seen the opinion of a single panel voting member change on the decision. Why? Because the stuff Chris has posted
is meaningless.