Fabien's open letter to the community

[CatPower]

If you throw 100 million at a project in a science where everything is about testing, of course you always win.

Why would anyone or any government spend 100mln on a chess program?

[Tom Barrister]

Hi, xann. I haven't seen you since the IOS days and Turtle and company. Good to see you're alive and well.

Wow!!!

What was your nickname?
Did you write a program?

I spent a lot of time on Othello.
Because of this, my approach to chess search was much more mature than say evaluation.
Thinking about it, it also affected evaluation: continuous game phases for instance.

For readers, <iOs> stands for Internet Othello Server (in the nineties).
Turtle is one of the programs I wrote during that period.

Fabien.

I was kingt on IOS. Didn't have a program, rating never made it much above 1400 (which is low for Othello there). I never got the hang of the game, could barely beat ant (one-ply Logistello) 1 game in 3, if that. I gave up about 1999, just about the time the game was considered solved as drawish with winning chances for White, and about the time the 16 piece random thing started up.

For those who aren't familiar with the Othello scene, no human can touch any of the top computer programs there without using a prepared line, and it isn't close. The problem is that humans did use prepared lines to win/draw. The 16 piece random game changed that. A world-class human would do well to stay within 20 pieces of the evaluation in a random game against the top programs. Logistello was probably the top banana, although it wasn't a big gap, followed by (in no order) Turtle , Hannibal and Zebra. All four of those won a tournaments each with the other three present.

[Robert Flesher]

.....the verdict is always "not guilty" as opposed to "innocent"............

I always thought that a court either convicted someone or cleared them of guilt.

Not under U.S. Law. As Dr. Hyatt has said, beyond a reasonable doubt is the standard of proof required for a conviction in a criminal (as opposed to a civil) case. It does not establish innocence… just that the “facts” and evidence of the case, presented by the prosecution, were not enough to lead a “reasonable person” to believe beyond any reasonable doubt that the accused had met or committed certain acts or elements that led to the violation of a criminal statute (hence a crime being committed).

Example from an Indiana Statute:
IC 35-43-4-2
Theft; receiving stolen property
Sec. 2. (a) A person who knowingly or intentionally exerts unauthorized control over property of another person, with intent to deprive the other person of any part of its value or use, commits theft, a Class D felony.

So you must (all terms defined by statute):
(a) knowingly or intentionally
(b) exert unauthorized control over property of another person
(c) with intent to deprive the other person of any part of its value or use

If a prosecutor cannot convince a reasonable person (twelve in this case) that a person committed each of those elements (beyond a reasonable doubt, not 100% doubt, just what is reasonable)… there is no conviction or establishment of guilt. Innocence is an entirely another matter.

The type of case we are talking is a civil case… again Dr. Hyatt has said this, one that uses a much lower standard of proof, that of “preponderance of the evidence.” Preponderance of the evidence is nothing more than a fancy way of saying: “more likely than not” or a greater that 50% chance. Put everything in the scales… if it tips slightly in one direction… that is “preponderance of the evidence.” It is the burden of proof I have argued before that should be applied in instances such as this (though with some of our programming experts, I dare say, they sound convinced to a higher standard, one exceeding “beyond a reasonable doubt” yet that is still not enough for many).

What I cannot understand is… why this community does not select individuals with certain programming qualifications who have researched this matter thoroughly to the point of giving an educated opinion… and then abiding by that decision. Would not that be preferable? The community “policing” itself? If there were 9 chess engine authors and you get a majority vote (and a brief, layman’s explanation why their opinion).

Graham I know you to be a reasonable man... would something like that, a "review board"... be acceptable to you? It's basically a rhetorical question and I certainly do not expect an answer (I am not attempting to “put you on the spot”). Something like that would be wise to consider for a time… and in this case with Fruit/Rybka it is essentially a loaded question.

I think that any case involving the issues presented here will most certainly fall under the same verdict ruling that civil law follows, that being the " balance of probabilities". These situations are almost always solved under tort law rules from my experience. Whereas "beyond a reasonable doubt" is the standard for the more serious criminal acts such as, crimes against a person ( assualt, robbery, theft, murder, etc, etc), or crimes against property. I am not sure how this works in other countries, but this is a brief explantion of how it works here in Canada. With tort law it will not be easy for Vas ( trust me I have vast experience here), as the balance of probabilities weight heavily against him.

I think you mean "preponderance of the evidence" which is, simply, that there is more evidence supporting one side than the other, but only a simple majority, not the "reasonable doubt" standard for criminal proceedings.

In my occupation I have spent the last 7 years in court hundreds of times. I was pretty sure what I meant, but perhaps it did not translate to my post. There is a clear difference in which different illegal acts are dealt with here in Canada. As I explained there are the "criminal" acts of a more serious nature ( see Martin's annual criminal code), these being summary, indictable, and hydrib offences. It is these offences that use the "beyond a reasonable doubt" for powers of arrest and then conviction.
( there are also provincial offences, but they do not concern with this comparision)

Tort or Civil law uses "the balance of probabilities"( also see, "burden of proof") this is the system in which the current Fruit-Rybka affair would be addressed here. The point being, it is MUCH easily to be found innocent when using " beyond a reasonable doubt" doctrine (as in criminal proceedings) as all that needs to be achieved is the introduction of "DOUBT". I can tell you from experience, lawyers are damned good at this!

[JuLieN]

@Diep
Nice to see you back, Vincent, and in what is obviously a great spirit!

[Robert Flesher]

.....the verdict is always "not guilty" as opposed to "innocent"............

I always thought that a court either convicted someone or cleared them of guilt.

Not under U.S. Law. As Dr. Hyatt has said, beyond a reasonable doubt is the standard of proof required for a conviction in a criminal (as opposed to a civil) case. It does not establish innocence… just that the “facts” and evidence of the case, presented by the prosecution, were not enough to lead a “reasonable person” to believe beyond any reasonable doubt that the accused had met or committed certain acts or elements that led to the violation of a criminal statute (hence a crime being committed).

Example from an Indiana Statute:
IC 35-43-4-2
Theft; receiving stolen property
Sec. 2. (a) A person who knowingly or intentionally exerts unauthorized control over property of another person, with intent to deprive the other person of any part of its value or use, commits theft, a Class D felony.

So you must (all terms defined by statute):
(a) knowingly or intentionally
(b) exert unauthorized control over property of another person
(c) with intent to deprive the other person of any part of its value or use

If a prosecutor cannot convince a reasonable person (twelve in this case) that a person committed each of those elements (beyond a reasonable doubt, not 100% doubt, just what is reasonable)… there is no conviction or establishment of guilt. Innocence is an entirely another matter.

The type of case we are talking is a civil case… again Dr. Hyatt has said this, one that uses a much lower standard of proof, that of “preponderance of the evidence.” Preponderance of the evidence is nothing more than a fancy way of saying: “more likely than not” or a greater that 50% chance. Put everything in the scales… if it tips slightly in one direction… that is “preponderance of the evidence.” It is the burden of proof I have argued before that should be applied in instances such as this (though with some of our programming experts, I dare say, they sound convinced to a higher standard, one exceeding “beyond a reasonable doubt” yet that is still not enough for many).

What I cannot understand is… why this community does not select individuals with certain programming qualifications who have researched this matter thoroughly to the point of giving an educated opinion… and then abiding by that decision. Would not that be preferable? The community “policing” itself? If there were 9 chess engine authors and you get a majority vote (and a brief, layman’s explanation why their opinion).

Graham I know you to be a reasonable man... would something like that, a "review board"... be acceptable to you? It's basically a rhetorical question and I certainly do not expect an answer (I am not attempting to “put you on the spot”). Something like that would be wise to consider for a time… and in this case with Fruit/Rybka it is essentially a loaded question.

I think that any case involving the issues presented here will most certainly fall under the same verdict ruling that civil law follows, that being the " balance of probabilities". These situations are almost always solved under tort law rules from my experience. Whereas "beyond a reasonable doubt" is the standard for the more serious criminal acts such as, crimes against a person ( assualt, robbery, theft, murder, etc, etc), or crimes against property. I am not sure how this works in other countries, but this is a brief explantion of how it works here in Canada. With tort law it will not be easy for Vas ( trust me I have vast experience here), as the balance of probabilities weight heavily against him.

I think you mean "preponderance of the evidence" which is, simply, that there is more evidence supporting one side than the other, but only a simple majority, not the "reasonable doubt" standard for criminal proceedings.

In my occupation I have spent the last 7 years in court hundreds of times. I was pretty sure what I meant, but perhaps it did not translate to my post. There is a clear difference in which different illegal acts are dealt with here in Canada. As I explained there are the "criminal" acts of a more serious nature ( see Martin's annual criminal code), these being summary, indictable, and hydrib offences. It is these offences that use the "beyond a reasonable doubt" for powers of arrest and then conviction.
( there are also provincial offences, but they do not concern with this comparision)

Tort or Civil law uses "the balance of probabilities"( also see, "burden of proof") this is the system in which the current Fruit-Rybka affair would be addressed here. The point being, it is MUCH easier to be found innocent when using " beyond a reasonable doubt" doctrine (as in criminal proceedings) as all that needs to be achieved is the introduction of "DOUBT". I can tell you from experience, lawyers are damned good at this!

[Ant_Gugdin]

Tort or Civil law uses "the balance of probabilities"( also see, "burden of proof") this is the system in which the current Fruit-Rybka affair would be addressed here. The point being, it is MUCH easily to be found innocent when using " beyond a reasonable doubt" doctrine (as in criminal proceedings) as all that needs to be achieved is the introduction of "DOUBT". I can tell you from experience, lawyers are damned good at this!

Hi Robert

You are right about the distinction between the civil and criminal law standard of proof (assuming the applicable law applies the same standards of proof as the US or the UK). However, the criminal law standard of proof may also be relevant here, as Vas may have committed a criminal offence under the law of Poland, where he is resident. Disseminating a derivative of someone else's work without authorisation (there is no authorisation here as there is no GPL compliance) is a criminal offence in Poland. The maximum sentence is three years' imprisonment. If the perpetrator "makes the criminal offence a regular source of income" or "manages" a criminal activity of this type, he will get a prison sentence of between six months and five years. Other Polish law criminal offences are also potentially in point. I pasted the relevant provisions of the Polish legislation into an earlier post on this thread.

If Fabien/ the FSF sue Vas (or perhaps Chessbase) for damages and/or the withdrawal of Rybka 4 --> Civil law standard of proof

If the Polish police prosecute Vas for the offence of disseminating a derivative of someone else's work, without authorisation (and/or another criminal offence) --> Criminal law standard of proof

[diep]

If you throw 100 million at a project in a science where everything is about testing, of course you always win.

Why would anyone or any government spend 100mln on a chess program?

If you'd use non-top secret classified learners/parameter tuners, let's say only stuff that's there out in the public, to achieve the same results at hardware from back then you'd need some billions rather than a 100 million.

In such type of projects they are not allowed to post the code even to show *how* the input works.

What all kind of programs share now is that even the connectivity to the learner is not there to modify parameters within the program. Each team shows up with their own silly excuses.

Why don't you ask them code towards the connectivity, as they can't afford to give that.

They'd get shot.

Vincent

[diep]

@Diep
Nice to see you back, Vincent, and in what is obviously a great spirit!

I am not back in computerchess at all. Too much money has been thrown at computerchess to beat guys who could not even afford testhardware at home.

It is not new to fund projects there. In 90s so many got funded. I remember also neural networks that got especially funded back then for learning; most of them had a cost of some thousands to millions each. That's all ok, as they didn't play a foreground role back then.

Accurate tuning now has become so important to computerchess, as it also fixes bugs, and increase in hardware allows now to get things done there, that the government dudes can play a foreground role while playing the Einsteins.

What happened here is much more than just fund a few projects.

What's tougher to prove, yet very obvious if you joined all events, is that specific engine and engine authors have been systematically sabotaged in tournaments, which statistically lost them many points.

It was 'coincidentally' always original engines with original authors who got sabotaged. Never one of the obvious N*SA guys joining in events, these guys were the guys profitting always, even if they joined with an elo 2000 engine.

It is difficult to do statistics with 1 occasion, but if you put in a line 20 of such occurances, then a systematic pattern is there, which already gives an error of under alpha 0.05 ( 2SD ), in court more than accepted as a proof.

In fact new medicines already get allowed with alpha 0.05 ( = 2SD), which delivers those manufacturers billions (you can really say this is total outdated way to deal with it from government side - that's another discussion, government still lives in 1950s).

You can be sure it drew a lot of attention world wide. Suddenly China got interested past few years, Brazil lately, and though i completely missed that chapter, it seems also Cuba tried to get their guys in.

[JuLieN]

*feeling scared, for some reasons *

[diep]

If you throw 100 million at a project in a science where everything is about testing, of course you always win.

Why would anyone or any government spend 100mln on a chess program?

There is a few valid excuses to give *some* funding, yet we must distinguish roles.

As you know from the wikileaks, USA has roughly 1200 intelligence agencies and 3 million people all together work there (the official numbers given there are different from 2.6 to 2.7 million to 3.0 million, by different congressional members so i hope you allow me to round it off at 3 million).

They doubled in size short after september 11, of course losing a lot of their personnel members there. Funding also more than doubled i assume.

Even the communistic party in Netherlands (called: SP = socialistic party, a direct marxistic party), they had in their 2007 election program an increase of 0.5 billion for the security forces, which is an incredible amount if you realize they are *against* having an army at all.

What do you do with all that cash?

First of all there is a number of valid 'see ya' guys who show up in tournaments. They have a good excuse to be there. Too much has happened there. Yet all they require is a few tickets and some hotel costs.

That's not the $100 million i refer to.

Vasik is excused there.

To give just 1 of the many examples excusing Vasik yet not the dudes who carried out the job of programming and wasting money to hardware:
It's not so long ago that Ahmad Altani left from USA back to Qatar, short before september 11. He was right under the nose of the USA, working at a military airbase. Really i didn't know that. It's the secrecy of the job that allows most dirt to happen. Had i known in 90s he worked at a military airbase, rather than the lame excuse he gave what sort of work he was doing, i would've immediately picked him out as someone who would've been a potential risk, as it was obvious to me he was a hardliner in religion.

Instead of that, he could use great hardware from others, Crafty and Diep and other software back in 90s ran on those Alpha's, cpu's not yet released, at the ICC server.

The evil part therefore is not the see ya guys who walk around; it is using that funding to build kick butt software, just because they can.

Instead of finding a method to do matrix calculations more accurately using a new algorithm that's lossless, i understand from conversations that currently a lot of problems in quantum mechanic get caused by round-off errors because of the floating point used, which let's the scientists use the slowish O ( n ^ 3 ) matrix calculations to solve the problem in that more accurate manner, they like to get more credits of course than that. So instead of helping their nation forwards, instead they join forces to fund an insane project like this.

Some guys like Fabien show a way to progress forwards and if someone else shows the fact it is possible, to just throw big hardware at it that eats on its own already dozens of millions in hardware, is of course something the non-brilliant working for the government can do as well.

The total amount of non-see ya guys that got funded is simply too big for this, as they did not do fundamental research. They only joined a contest, and were obviously paid to do so, as secret information and secret software was used to get things done.

What you then typically get is dozens of consultants from the health-IT and bioinformatics, so guys who really have no clue on computerchess, let alone game tree search, who join in, just to catch their rate of 75 euro an hour up to 168 euro an hour. Most are around 100 euro an hour.

Oh yeah, they need also a program to show up at a tournament. Ring ring, can i get a clone for myself? Of course they ring up everyone within government, but not the chessprogrammers. Some of them do ring up a few chessprogrammers. I also had a few on the line, but they definitely didn't want to give more than a 500 euro for a 'kick butt' clone. No business with me was done. Other chessprogrammers reported (i heard this from 3 different chessprogrammers) for very low fees. Around $30 to 30 euro an hour or so for a limited time usually.

So these crap people with their clones, who also pollute computerchess, show up *because* of the smoke already created by N*SA.

It has also very bad side effects.

As government even has to manipulate the tournament directors to not catch those guys as full blown clones, whereas even a kid on the street recognizes them as an actor and clueless.

If you obviously and willingly manipulate the entire field in such a manner that you are in 100% control over the field, all this with government funding, then that's what i call a very evil way to get things done.

For your information i have been in a tiny energy commission with respect to high voltage powerlines. Based upon knowledge i have from that i can assure you that when in 6 years from now USA must import the full 19 million barrels of oil daily into USA, which will have a cost by then of maybe even more than a trillion dollar a year, that in short term things will go very bad in USA. The military will be used for basically everything.

That's what happened in Greece also. Past 50 years, the national spendings to military were roughly 50% of the national income. A situation you cannot maintain. Total bankrupsy.

Yet despite that this energy problem is the biggest problem threatening USA in short term, if you would project the manipulation there onto a small world like computerchess, manipulation in computerchess by government has been a lot bigger.

Just because they could afford to do it.

100 million is peanuts for governments, if you realize how much money they waste anyway.