Wednesday, January 25, 2006

Iron law of oligarchy

Its a pretty interesting concept. Basically put its the idea that any system regardless of how it is started will migrate to an Oligarchy.

Be it polical, social, or even now networks and machines.

Okay so here's the scineros:
a) Start as a Monoarchy or Dictatorship. As the system grows the monoarch or dicator will gradually start delegating tasks and decision making to those he assigns simply because there are too many decisions or tasks or simply wants a break. In turn those will start to specialize and become essential to the functioning of the government and will at some point challenge the leader. This then is the begining of the Oligarchy. Oligarchy doesn't say that all the few share the power equally, only that a few control the majority of it.

b) Start at a Democracy/Republic. Either everybody elects a leader or votes on every issue. Eventually people get sick of this and will elect representatives to make those decisions for them. Eventually those representatives can institutionalize themselves though the appathy of the people. And an Oligarchy is born.

Whats facinating to me isn't the nature of people, its how it even occurs in machines and networks. My best guess is its only related for our natural tendancies as engineers of these systems to put them in a hierarchy. And in general as systems get larger our ablity to create the master computer is limited to our technology we end up reverting to several imply massive computers to manage the rest of the system. Each of those is specalized to due certain tasks. While not quite to the point of making network changing decisions (some systems are very close to being able to detect intrusions and self heal the last time I checked) because we have not given them that sort of programing. I would have no doubt in time that we could have at least moderate AIs running our large computer systems.

Of course the peer2peer and distributed processing movements could end up changing the hierarchical approach so we just to wait until we have more decision making ablities built into computers.

Editors note:
I bet you could make this a Ph.D. thesis. First develop two computer only models. The basis of these models are computers need to make "important" decisions or simply calculations. Give the programs on these computers the ability to know what the performance of the others in the group are. Allow each one to decide to do the decisions themselves or allow them to delegate the task. First start with a small group of equal performing computers. Then start inserting more powerful computers and less powerful ones. See if they move to an Oligarchy. Then try the other massive computer first making all the "decisions" and start adding smaller computers it can chose to delegate to, then add even less powerful ones.

0 Comments:

Post a Comment

<< Home