Computer chess breakthroughs and imitations

This new series of articles, by Dr. Søren Riis, here, and here, and here is of general interest and does not require chess knowledge.  They are an excellent case study in innovation, IP, reverse engineering, incentives to copy, market leadership, and other currently important concepts.  Excerpt:

This program was immediately thought to be a very close derivative of Rybka because its solving of test positions was extremely similar. But, beyond the objective measures of similarity testing, Strelka had to have been a reverse-engineered Rybka derivative because, at the time, a new program of such strength and manifest similarity in its playing style could hardly have come from anywhere else. Thus a very public precedent was established: someone had reverse-engineered a closed-source program with impunity.

Comments

Strelka was a trap, designed to catch the Rybka programmers and it worked perfectly.

It was a Fruit clone, disguised as a Rybka derivative. They waved it in Rajlich's face until he could ignore it no longer. But when he finally accused it of plagiarism, what he really did was incriminate himself - claiming that Strelka was a Rybka 1.0 clone, when it can be shown that Strelka was (also) a Fruit 2.1 clone. If Strelka=Rybka and Strelka=Fruit, then ....

Comments for this post are closed

There is a lot of loose language in computer chess because there is a continuum between a perfect clone that has simply been relabeled and a program that used ideas from other programs and reimplemented them. Thus most discussions of program "copying" are really not about source code copying but rather different degrees of taking ideas and algorithms--in Strelka's case, by reverse-engineering Rybka.

As mentioned in the article, Rybka's author has credited Fruit's author on multiple occasions for his contribution to computer chess programming. There is no question Rybka's author learned a great deal from Fruit: he admits it. However it would be a travesty to say Rybka is a Fruit clone or even a Fruit derivative as is redundantly explained in the article: different evaluation, different search, different board representation, far less correlation between moves played than is in evidence when comparing several programs that have emerged in recent years that benefited from seeing reverse-engineered versions of Rybka.

If a product is functionally 10% better (approx. 300 ILO points) isn't discussion of how it's only a trivial clone somewhat off the mark? Without doing something original how does Rybka gain 300 ranking points.

Comments for this post are closed

Comments for this post are closed

A certain irony here, because a man with the same last name, Bjarne Riis, was one of the great cheaters in bike racing.

http://sports.espn.go.com/oly/cycling/news/story?id=2882380

I do not know if they are related in any way.

Comments for this post are closed

Tyler - I'm reading an interesting book that's highly related to Chess A.I. right now: The Most Human Human, by Brian Christian. Well worth checking out.

Robin Hanson had a diavlog with Christian, which he links to and discusses here.

Comments for this post are closed

Comments for this post are closed

In my opinion the accusers' have four main arguments (at the layman level that Riis argues):
1. without copying the code, Rybka' programmers would have had less time to focus on those tasks that improved the performance - it helped them beat the competition
2. initially, the increased performance wasn't due to improvements in how the computer thinks about chess, instead were concentrated on how the algorithms were implemented (better use of the new CPU's capabilities - 64bits, etc)
3. Søren Riis is actually a moderator on one of the Rybka forums, and associated with the company that distributes this software - hardly an impartial mind
4. (and this is the one I find the most disturbing) - there was no paradigm shift in how computers play chess - indeed it has improved in the last 5 years, but the paradigm shift story is a mighty exaggeration

Comments for this post are closed

A rebuttal to the series by Riis for those that are interested.

http://www.chessvibes.com/sites/default/files/pdf/Riis3_Mark.Watkins.for.ICGA.pdf

Comments for this post are closed

Comments for this post are closed