Scattered de/centralization notes

why is the image of rapid prototyping, or high iterations selction etc, not evolution?  it has something to do with decentralization

how is evolution not decentralized or independant of decentralization?



basis behaviors, modules and stable subassemblies are not fundametnal to conventional enginneering, they are also used in decentralized systems engineering.



modules are different from stable subassemblies (or whatever.  In the case of an orVocabulary must distinguish between a 

in the case of an organism with five parts that can each perform five functions, and five function that must performed for a part or 'related' set of parts to survive, we have two polar ways to distribute the tasks.  

The first is to have each part each perform all five tasks.  In this example, we essentially have five different organisms.  if they are coupled and sharing resources for a time, but get separated, each individual part will continue to function and to survive.

The other pole is to have each part specialize (devote all five of the jobs it can perform) to one task.  Have all five parts couple, so that they are sharing resources and collectively, each pat is recieving teh benefit of all five services.  presumably, this collection, this multicellular organism, is better at competing for resources than any of the individual parts from the example above.  But if one part of this organism gets separated from the others, than all five parts will die, because no part is finding all five necessary functions filled.  

The organism from the second example each part reduced its complexity.  The complexity of the parts in the first example is higher than that of the parts in the second one.  This is the same as saying that the number of states of each 'autonomous' part is higher than the number of states of each restricted or specialized part.  But the complexity of the collection of the five parts is higher in the second example.   By restricting their own complexity, they made complexity at a higher level possible?  What is the evidence that the whole

Hello Yaneer, I made a toy model that has some exciting explanatory power.  I've put some effort into explaining it clearly, so please give it a fair read.  It has cells with parts and they can share resources.  It is a way to understand changes in complexity over scale in the context of cooperation, and also, with very few parts, offers valuable clarification of ideas like robustness, modularity and your plant and animal distinction.  So:

There are 5 cells.  Each cell has five parts and manufactures one unit of chemical in each part.  Each cell needs to have 5 different chemicals at any moment to survive.  If a cell is not receiving a unit of all five different chemicals, it dies.  

In the simplest case, each of the five cells is using each part to produce each chemical.  They are floating about on there own and producing all they need to survive and are entirely self sufficient.

Now we will complexify the cells just a bit.  Let them share chemicals.  This makes possible (but not practical) an important new configuration: We can make it so each  cell is specializing in one chemical.  Cell One makes five units of Chemical One.  Cell Two makes five units of Chemical Two, etc.  All cells are sharing, so each cell is receiving the five chemicals that it needs to survive. 

I noted that this is possible but not practical because if we separate one cell from the other four, than all five die.  The removed cell is receiving five units of only one chemical and the other four are receiving all but that chemical.  At this step of my model, the collective arrangement offers no advantages over the 'individualist' arrangement . 

Now let us introduce economies of scale.  Within one cell, if Chemical One is being produced by two of the cell's parts, it gets four units of that Chemical One.  If it is produced by three parts (3/5 of its manufacturing capacity) it produces 9 units of Chemical One and so on.  The benefit can be any kind of greater-than-linear growth, i chose n^2 for this model where n is number of parts devoted to manufacture.  Lets call the exponent of this function the "economy".  In this case the economy is 2.  In the starting case, it was 1.  

The economy is important.  Where the economy was 1, there was no difference between the collectivist and individualist arrangements.  With an economy greater than one, the collective arrangement is more 'complex' than the individualist arrangement.  Where five individualist cells collectively produce 25 units of chemical, the cooperating cells collectively produce 125 units of chemical.  With my imagination and knowledge, I can only justify the use of the word 'complex' by imagining an environment where surplus chemical allows survival in more hostile conditions.  But that is an ugly thing to build into an otherwise simple formalism.  How else can I call the latter more complex?  Regardless:

Lets define efficiency (trivially, for now) as the number of units produced by a subset of cells sharing resources.  We are now looking at a tradeoff.  Where, as individuals, the cells are robust to separation, the cooperating 'collectivist' cells are more efficient than individualized competitors.  At intermediate levels of cooperation, we will see robustness and efficiency slide past each other.

The economy exponent provides the incentive to scale up.  It provides an incentive for higher scale organization.  Not only does this exponent control the extent to which  cooperation gets rewarded, my preliminary poking around indicates that it also controls how 'plantlike' or 'animallike' a collective of units is.   Really!  But I won't go into the formalism that shows it, that would take up too much space without any pictures.

So, in summary, this formalism provides a way to analyze changes in complexity of a simple system over changes in scale.  It also provides a way to analyze plantlikeness and animallikeness.  It also suggests different kinds of robustness and nuances the definition of the word modular.  Both of these concepts are far too butchered, and the model shows some of the nuance in them.  

Also, my discussion only looked at the extreme ends of the spectrum, full collectiveness and individualness, but the intermediate points on the spectrum are ripe.  My preliminary work suggests that plantlikeness and animallikeness are meaningless distinctions at economies of one.  it is a simple and neat discovery, and only the beginning.  




Pictures and person to person communication become much more useful than up to this point. introduction.



With an exponent of one ( the introductory case ) lets look across the spectrum at what plants and animals are.  The individualist extreme is trivial, the plant animal distinction is meaningless.  At the cooperative extreme, the distinciton is also meaningless.  Any part removed will kill the system, and that isn't plant like or animal like.  
At intermediate states of cooperation, the only configurations that don't kill all five cells

If you play around with intermediate states of cooperation and an exponent of one, 'animals' and 'plants' are the same trivial thing


Remember your definitions of plant and animal?  where you sever a plant and both parts are still alive?  That is possible as an intermediate state of this 


before going to the next step lets dfine a bit better the notion of a chemical, so that chemical is conserved.  each cell can produce five units of up to five different kinds of chemical, and each cell needs one unit of each chemical to survive.

with this formalism, we see that all intermediate forms of cooperation are also vulnerable.  If a 'colony' is defined as a collective made entirely of self sustaining parts, a 'collectible' (fragile) as entirely of completely specialized parts, an 'animal' is defined as a group where removing a part (can?) leave some subset of teh collective alive and a 'plant' as a case where separation can leave both parts alive. I propose that plants are impossible via this scheme unless they are colonies.

there is a level of formalism that i am leaving out here to spare you complexity while I introduce the model, but I have proper definitions of 'vulnerable', 'chemical', 'survive' etc.

We can also note that between these two arrangements are arrangments where separation of a cell would kill only that cell while the other four remained intact.  Also note that this arrangement has as many states as the case where all five are connected, but each is producing its own chemicals to survive.  five units of five different kinds of chemicals are being produced in both cases; 5^5 possible states of both configurations.


robustness to what?


how do you count states? what are you counting?




what is the correct extent to which we should learn to think in 'life' or constrain life tand make t think in our langaugeinstead?





I just took a course exploring current work in engineering decentralized systems and found an interesting clash, mostly within myself, as to the best way to make such engineering tractable.  I work at for an educational non profit interested in helping people deal with complexity.  How do we manage systems that are more complicated than one person can understand?  Such systems exist, large corporations, governments, societies and living systems, and in cases where such systems are given a leader, we have to understand that this person has very litle clue what they are doing and can do.

NECSI, my employer, is interested in developing tools that help such people understand and predict the consequences of such interventions.  We are also interested in the tools behind designing such unmanageable systems, for all their benefits, in a way that they accomplish tasks planned by their designers.  You can see it as a new kind of engineering, and it involves a new kind of thinking, or a new language to think in.  Much of my thinking in the course was from the angle, 'How do I think in the language of this system'?  

This contrasts with the approach of the majority of my classmates, and that of the instructor.  They were all engineers of one sort or another, and I found that the questions and directions worked towards understanding 'How do I conceptualize this in terms of the engineering process'?  Essentially, 'How do I make the system work in a way that I'm used to thinking?'.  This would take the form of simplifying the behavior, finding independent 'basis' behaviors that can act as modules and biulding