Robert, Do you know what Occam's razor is?
Einstein originally believed that the universe was static. When this didn't fit his observations he invented the "cosmological constant", which he considered one of his biggest blunders. If we are going to continue to discuss this, then if you have a theory that has no yet existing evidence (such as sudden diminishing returns) then YOU present evidence. This is called the burden of proof. If you accuse someone of some crime, it isn't up to them to prove you wrong, it is up to you to prove your assertion. You might believe that it's up to me to prove that there isn't a sudden diminishing returns affect, but you have it backwards. It would be pretty unreasonable to continue to throw assertions at me that I have to prove - YOU are the one that needs to prove your assertions. I think there are little green men living about 20 miles down below the surface of the earth. Prove me wrong! See, you can't so there! The facts are that Mogo is strong, it's CLEARLY stronger on big Iron and they went to a lot of trouble getting big Iron based on that fact. We also observe that the good programs do better against both humans and other programs. Do you have something we don't know about that is more than a hunch? - Don On Sun, 2008-08-10 at 16:46 -0400, Robert Waite wrote: > > I don't know how you can say that. The empirical evidence is > > overwhelming that this is scalable in a practical way but more > > importantly it's been PROVEN to be scalable. If you throw the word > > > "practical" in there then you are no longer talking the language of > > mathematics, theory and proofs so please don't ask for a "proof" of > > "practical" scalability, it makes no sense. > > I would really like to see what paper you are referring to. Do you > mean "Bandit based Monte-Carlo Planning"? Please post the name of the > paper which you are referring to. I do not think that the empirical > evidence is overwhelming that it is scalable in a practical way for > the problem of beating a human. > > "PROVEN to be scalable"? Big deal... isn't an algorithm that does an > exhaustive search provable to be scalable in the same sense? The fact > that it is proven to be scalable as the sample size increases to > infinity does not help the cause. The only thing that helps is the > rate at which it approaches "perfect play". How is this different from > exhaustive search with regards to being proven to be scalable? > Exhaustive search is scalable in that I could give it all the memory > and time it wanted. And it would approach a finite amount of memory > and a finite amount of time. > > Complexity theory is based on math and it does address "practicality". > By using the word practical.. I am not jumping into mysticism. I feel > that the proof that you offer does not help us in a practical sense, > at least in a rigorous mathematically proven way. > > > We are of > > course talking about the issue of scalability in a practical game > > improving sense. > > Okay.. so where is the paper that correlates the speed at which MCwUCT > approaches perfect play with the ability to play a human? They seem > unrelated as of yet. > > I think it's very likely that the > diminishing returns curve will be very very gradual for a long time to > come, well beyond the point of achieving the top human levels. > This is conjecture... and it does not relate with MC methods being > proven to be scalable. It's a gut feeling.. just like many feelings I > have about go. > > > When our realities don't match > > our belief systems, we balk. > > If you take them off the > > pedestal, you can think more rationally about it. > > I don't think that represents my feelings on the subject. My gut > feeling before the match was the learning machines and further > advanced in AI would be needed to solve the problem. This was from a > sense of the potential intractability of go. I could very well be > wrong. > > It's obvious that you could recreate a brain since it is made of a > finite amount of matter. So I have no mystical attachments to the game > of go. I just think we have not proven yet that number crunching > methods are viable alone. A more heuristic approach could still be > needed. Mogo does use MC methods to play.. but it does have a few > heuristics to help judge important trees. Will number crunching > methods be enough alone... or will there be a need for much stronger > heuristics to trim the tree? I don't think we know yet. > > > > > > > _______________________________________________ > computer-go mailing list > computer-go@computer-go.org > http://www.computer-go.org/mailman/listinfo/computer-go/ _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
