[computer-go] Decision-making methods in a distributed "swarm" of "light-weight" processors.

[Aidan Karley]

        I tried to make the "Subject:" relevant to the article, but also to raise 
the interest of Go programmers, particularly those using parallelised systems.

Article at www.americanscientist.org/template/AssetDetail/assetid/50768 and in 
the over-the-counter editions where I saw it. 
Abstract :
> Group Decision Making in Honey Bee Swarms
> When 10,000 bees go house hunting, how do they cooperatively choose their 
> new nesting site?
> Thomas D. Seeley, P. Kirk Visscher, Kevin M. Passino
> One of the turning points in the life of a honey bee colony is when  a 
> queen bee bequeaths her hive to her daughter queen, takes half the worker 
> bees and goes off to start a new nest. The departing bees' process of 
> deciding on a new home seems to take some time. With careful research, 
> Thomas Seeley and his colleagues have uncovered how a swarm comes to 
> a decision. It's not a democracy, exactly, but rather a matter of reaching 
> a threshold as bees endorse a particular site using their "waggle dancing." 
> For a group of about 10,000 bees, several hundred scout out nest sites, but 
> it takes the build-up of just 10 to 20 bees at a site before the swarm 
> starts to move to that location. Through experiments and mathematical 
> modeling, Seeley's group has shown that the bees' method is best at 
> balancing the need to find a home quickly and choosing an ideal nesting 
> site.
>
       I don't have a subscription to get the electronic full text, but I'd 
suspect that at least a few other people on the list would have met Am.Sci. 
before and be able to vouch for the quality of it's articles.
       
       What interested me is the way that a number of lightweight (in processing 
power, and literally!) evaluating units (bees) could quickly and effectively come 
to an opinion as to the optimal move to make in their area of exploitation. 
Further, the method that appears to be used involves repeated sessions where a 
particular processing unit (bee) re-visits a site being evaluated, where the bee 
re-assesses it's opinion of the quality of the site. 
       This strikes me as somewhat similar to the way that I play - coming up 
with a list of possible locations to play, then repeatedly looking at possible 
moves on from each possible move, discarding some moves and re-iterating until I 
have just one move which I've followed on most deeply as being my best choice.
       
       Well it strikes me as having some relevance in an AI sort of way to Go 
programming. Whether it's something that could be an effective evaluating 
algorithm for pruning move selections down to "the move", I don't know. But I 
think it's at least worth a look.
       
-- 
 Aidan Karley,
 Aberdeen,  Scotland