Paradigm shifts

[Gian-Carlo Pascutto]

A train of thought I wanted to post in the DS thread, but which is probably worthy of a separate topic.

I see Ippolit in the same way as Fruit. It creates a new gold standard for engine development. If you write a new engine, you should study Ippolit inside out and consider it the reference of how to do things. The same thing happened with Fruit. The result was that any new engine was on average several hundred ELO stronger than was the case before Fruit.

It also created the unfortunate situation that any new engine looks a lot like Fruit. There was a thinning of original development and ideas. (Not sure how to say that properly in English)

I am sure that Ippolit will accomplish the same.

Maybe this is a good thing, and it's a kind of natural selection of ideas. But it's also a fact that any alternative approach will have a formidable baseline to meet before it can be considered. Without the Fruit or Ippolit sources out, there was more breeding ground for new ideas.

Computer go was going in the wrong direction for years, it took some very stubborn people to go the other direction for 10 years before they could meet the gold standard from the wrong direction and leapfrog over it. Now the opposite is happening: if you're not an UCT-like Monte Carlo with patterns playout based program, you're too weak to be relevant. But your approach might be the right one.

The chess paradigm has been to make an as fast as possible alpha-beta searcher with selectivity and heavy leaf pruning, and as much evaluation is practical. This has been remarkably stable over the last, say, 20 years. The extremities to which the approach has been pushed are amazing.

I think there was a paradigm shift when magic bitboards appeared. They are efficient enough that they make any other approach just look wrong. However board representation is not that defining for program strength. (This might be quite subjective)

What are, in your opinion, the odds we will see another paradigm shift? What might be a trigger?

I have some hope big manycore cpus (much more than 100 cores) might cause it. Maybe someday someone makes a workable, strong GPU program. I believe that to accomplish that, a paradigm shift is needed.

[Rolf]

A train of thought I wanted to post in the DS thread, but which is probably worthy of a separate topic.

I see Ippolit in the same way as Fruit. It creates a new gold standard for engine development. If you write a new engine, you should study Ippolit inside out and consider it the reference of how to do things. The same thing happened with Fruit. The result was that any new engine was on average several hundred ELO stronger than was the case before Fruit.

It also created the unfortunate situation that any new engine looks a lot like Fruit. There was a thinning of original development and ideas. (Not sure how to say that properly in English)

I am sure that Ippolit will accomplish the same.

Maybe this is a good thing, and it's a kind of natural selection of ideas. But it's also a fact that any alternative approach will have a formidable baseline to meet before it can be considered. Without the Fruit or Ippolit sources out, there was more breeding ground for new ideas.

Computer go was going in the wrong direction for years, it took some very stubborn people to go the other direction for 10 years before they could meet the gold standard from the wrong direction and leapfrog over it. Now the opposite is happening: if you're not an UCT-like Monte Carlo with patterns playout based program, you're too weak to be relevant. But your approach might be the right one.

The chess paradigm has been to make an as fast as possible alpha-beta searcher with selectivity and heavy leaf pruning, and as much evaluation is practical. This has been remarkably stable over the last, say, 20 years. The extremities to which the approach has been pushed are amazing.

I think there was a paradigm shift when magic bitboards appeared. They are efficient enough that they make any other approach just look wrong. However board representation is not that defining for program strength. (This might be quite subjective)

What are, in your opinion, the odds we will see another paradigm shift? What might be a trigger?

I have some hope big manycore cpus (much more than 100 cores) might cause it. Maybe someday someone makes a workable, strong GPU program. I believe that to accomplish that, a paradigm shift is needed.

How can you think this way? Please consider. You stated that you hesitate to publish Sjeng 4 because of the stealing operation against Rybka. And two hours later you salute the coming of a new gold paradigm with the inventions of Vasik?

Please make no mistake. Vas stood on Fruit in the beginning, but Fabien had left already and when told about Rybka he was amused and had no idea about counteractions at all.

Now we all know that Vas stuff was stolen and that further clone variations have no chance to survive because they dont play the Wch. Nor are played on servers. Sjeng 4 will play, no?

Perhaps you wrote tongue in cheek?

[Matthias Gemuh]

A new paradigm shift could be evaluation functions that tune themselves automatically, getting truely better from game to game, and maintaining only weights in a small fixed-size file.

Matthias.

[Bill Rogers]

Please make no mistake. Vas stood on Fruit in the beginning, but Fabien had left already and when told about Rybka he was amused and had no idea about counteractions at all.

Please excuse me for interjecting a comment here but I was here when Vas first made the statement the he had examince Fruit. He did not say that he copied even one line of code from it only that he had looked at the code. This same type of thought has been misquoted by dozens of pepole ever since Vas made his original statement.
People here are so quick to judge it makes me wonder to the moral validy of the majority of those who actually read and post, ie., are they all thieves of one kind or another?
Bill

[Rolf]

Please make no mistake. Vas stood on Fruit in the beginning, but Fabien had left already and when told about Rybka he was amused and had no idea about counteractions at all.

Please excuse me for interjecting a comment here but I was here when Vas first made the statement the he had examince Fruit. He did not say that he copied even one line of code from it only that he had looked at the code. This same type of thought has been misquoted by dozens of pepole ever since Vas made his original statement.
People here are so quick to judge it makes me wonder to the moral validy of the majority of those who actually read and post, ie., are they all thieves of one kind or another?
Bill

That was what I meant. He said he actually understood the code and wrote his own. I dont know if you knew that even Bob saw in Vas statement that it were his Rybka an admitting that he had himsel taken Fruit code because the anonymous Russian (?) Strelka "author" had declared that he based Strelka on Fruit. You see what this is basically?

IMO it's sort of blackmail. It's still beyond me why Bob bought this line. But as you probably know I have no experience in programming details so that I simply cant contradict Bob. All I can say is that the chain above doesnt prove anything for me. Because the Strelka guy isnt real in our understanding of legal authenticity. IMO.

[jdart]

We don't know the limits of what is possible in computer chess. The cluster based programs are pushing the limits in terms of NPS and overall depth. And the searchers are definitely getting smarter in terms of efficient pruning that doesn't reduce tactical strength, although I'm not sure at this point we have a new standard set of algorithms for that - we have a lot of variants that all perform reasonably well (Stockfish, Toga, etc. plus what we deduce is in Rybka).

--Jon

[M ANSARI]

Well if Vas keeps improving Rybka he can always stay ahead of the curve. Obviously if he finds that hackers can disassemble his code faster than he can release it then this could make him give up, but last I heard from him he still has a very positive outlook on the future. So while the stolen code of Ippolit might close the gap, I think Rybka is still ahead of the curve as the cloners will always be playing catchup. Most likely Rybka 4 will have better security and more difficult to crack.

With regards to other technologies out there, obviously if you can figure out how to code a chess engine to take advantage of massively powerful GPU's that would be a big plus. You can put 4 fast cards in Quad SLI today. Also cores are being added to CPU's very quickly with 100's of cores being available within a few years. If someone can take advantage of these technologies then obviously he would have a huge advantage.

[Gian-Carlo Pascutto]

A new paradigm shift could be evaluation functions that tune themselves automatically, getting truely better from game to game, and maintaining only weights in a small fixed-size file.

Matthias.

I think this would reach a diminishing return quickly, although I've been surprised by Stockfish.

Better would be a way to discover the features automatically.

[Gian-Carlo Pascutto]

although I'm not sure at this point we have a new standard set of algorithms for that - we have a lot of variants that all perform reasonably well (Stockfish, Toga, etc. plus what we deduce is in Rybka).

I think these all look very much alike. There's implementation differences but the general structure is identical. (just like any strong program produced in the last 10 years)

[Gian-Carlo Pascutto]

How can you think this way? Please consider. You stated that you hesitate to publish Sjeng 4 because of the stealing operation against Rybka. And two hours later you salute the coming of a new gold paradigm with the inventions of Vasik?

There's no contradiction. It's unfortunate but the cat is out of the bag, and it can't be stopped any more.

Any really strong new program is very likely to look a lot more like Rybka. This happened with Fruit and it will happen again.

Now we all know that Vas stuff was stolen and that further clone variations have no chance to survive because they dont play the Wch.

There's been several clones in the WCCC. The ICGA just cares about whether they collect an entry fee or not.

Sjeng 4 will play, no?

I'll play any tournament which I consider interesting and where the organizers don't ask me to pay a bribe.