More Theory

[Kmenex]

Who here has read the book "a new kind of science" written by Stephen Wolfram?

He seems to present an argument which leans favorably towards moving slowly out of the idea of the continuum.

I wanted to use this page to talk about the nature of discrete elements of time, space, substance, whatever, and the modeling of systems. I'll be posting some crazy ideas surrounding space and time in relation to CA systems that i have been trying to create and formalize and i am interested in what other people's opinions are towards modeling and constructing discrete abstract systems. Pardon me if i have overlooked many features of the description, but i fail to see any actual "model" of a complete COSA operating system. Because of this i do believe that there is still much value to be had in terms of flushing out a more tangible abstract model of a COSA system. Considering that a COSA system does not yet exist we are faced with the dilemma covered elsewhere in an attempt to implement the operating system in a virtual machine. I am no programming expert but i do feel that "adapting" COSA to the current language and processing schemes will not necessarily allow a smooth transition to a "natural" COSA scheme (one where the chip architecture is designed for such). It will give much insight no doubt as to how the structure will look, but i do believe that it is important to also have a "pure model" of how the system operates and what it is logically composed of.

I'm not a logician, i'm essentially a fool who knows how to operate some nanoscience tools + do some nanostructure synthesis. Regardless, i'm really interested in signal transduction and the design of "functional structures", and i think that there are so many exploitable physical, electrical and optical properties at the nanoscale that we can use to ponder and create new "functional element" designs and systems... Remember that it is just now that optical elements are seeing a pathway for integration with current silicon.

[harry666t]

Pardon me if i have overlooked many features of the description, but i fail to see any actual "model" of a complete COSA operating system.

Yes, there is no actual formal description of COSA, only Louis's prose. The ideas are constantly under refinement (mainly concerning the ways to implement the actual system), but the core isn't seeing any dramatic changes (maybe except for reactive data connectors, I believe Louis has first introduced the idea in this blog post: http://rebelscience.blogspot.com/2009/0 ... art-v.html).

Of course, this doesn't mean we wouldn't benefit from such a formal description. Actually, if you have the necessary skills & mindset, and could spare some time, maybe you could be the one to start the sub-project.

i do feel that "adapting" COSA to the current language and processing schemes will not necessarily allow a smooth transition to a "natural" COSA scheme

Nay. A model is a model, a byte is a byte, a logic gate is a logic gate. It will be "just" like all other transitions to a different architecture (except for a complete paradigm shift). Given that the backing execution environment is implemented properly in both places, what could possibly go wrong?

Now, the only question is: can we properly implement COSA in both software and hardware? Right now I'm devoting some of my free time to working on the software side, so it should give us a taste of the answer (and I see no reason for it to be a "no").

i'm essentially a fool

Don't talk of yourself like that. If you won't respect yourself, then who will?

[Kmenex]

Nay. A model is a model, a byte is a byte, a logic gate is a logic gate. It will be "just" like all other transitions to a different architecture (except for a complete paradigm shift). Given that the backing execution environment is implemented properly in both places, what could possibly go wrong?

I see. Lets hope then that it can be implemented properly.

can we properly implement COSA in both software and hardware?

That is the question, and that is why i emphasize constructing the abstract model because it will help in troublshooting and classifying how differnt parts of the system should be connected. I think that a graph-centric approach towards modeling the COSA system will prove valuable in the long run. Much to say, going to work on my Cellular Automata post.

[sixwings]

Kmenex, welcome to the COSA forum. We'll be looking forward to your ideas.

Please note that I, being the incorrigible rebel that I am, have a bone to pick with traditional approaches to software design. I believe in trial and error design. That is, I don't believe that a design should be completely specified before the implementation. This may be necessary in other engineering fields (less and less so due to powerful simulation tools) but I don't think it is advisable in software. In fact, I am convinced that software design and implementation should be combined into a single step or process. In fact, it is a never ending process. I believe that we are all designers and that new ideas for COSA, especially the look and feel of the visual studio, will pop up during implementation. We will either adopt them, refine them or reject them. This is the reason that I resist the temptation to sit down and flesh out a formal COSA specification. I think it's a waste of time.

[Kmenex]

Your words are always inspiring.

To hash out a system at this point is indeed a waste, for there is so much more to be looked at before that can even be done. And never in the course of emergence did the final state appear first. No... generally when we design a system we are already working with "components" whos forms are known and describing the end result is easy. When we are instead researching and developing, as is undoubtedly the case i see here with COSA, then we cannot proceed at more then a step in time. However, after we have taken many many steps then our familiarity with COSA both abstract and tactile (think perceptive pixel, think programming interface) will be such that we can "flush out" full systems quite easily. I share your sentiment of the trail and error method, for i am a nanotechnician who can barely solve a differential equation, yet alone model a molecular system with quantum effects in physics. Still, i am able to access closer and closer to the quantum domain with tools... Looking (ala wolfram's search) for interesting patterns.

I am still of the perception that theory should be pursued. However, it should be pursued via trial and error, whereby results from "physical" trail and error are considered in the construction of the abstract case. A solid graph-centric base can be used for the starting elements, where the actual COSA structure is represented by the graph grammar that constitutes the rules in the dynamics. Some people seem to think that all possible "abstract model systems" have been created, and that the "base elements" of mathematics are already found and described.

Soon i will make a post here talking about a Cellular Automaton system with interesting temporal update schemes and spatial connectivity.

[Kmenex]

COSA as a 3-d cellular Automaton

I have been interested in modeling discrete system in both finite and infinite domains with small collections of rules for quite some time. I was initially inspired in 2006 by Stephen Wolfram's book, "A new kind of science", in which he went through great length to show that simple programs can lead to extraordinary complexity which is essentially non deterministic. My interest currently revolves around graph-centric methods for modeling self assembly of forms and modeling and describing what i call "active rule space" in a special 1-d cellular automata system.

Recently i have been gaining increasing interest in 2-d and 3-d cellular automata and indeed i must say that when i think of COSA i imagine a 3-d cellular automaton. However, Cellular automata are generally spatially fixed, whereby the connectivity of cells is determined by some spatial matrix. Graph-centric models on the other hand have connectivity which is not dependent upon the "visual" or spatial layout of the system, and thus are much more efficient as abstractions for coding. The problem with graph-centric models is that they generally only deal with connections between nodes, whereby allowing the node to have both state and connectivity requires extra structure that can make things become unnecessarily complex. So it's a good part, but not the whole picture.

I am interested as to what people think of what cellular automata can do for modeling the COSA system. If anyone wants to learn more about the active rule space let me know. If anyone wants to check out the awesome work done by Eric Klavins at the University of Washington revolving around graph grammars for modeling and creating programmable matter, check out http://soslab.ee.washington.edu/mw/index.php/Main_Page and http://www.molecular-programming.org/

[Kmenex]

Links

I will use this section to post various links which are pertinent to the discussion of discrete parallel systems which produce useful global behavior.
----------------

check out Stephen Wolfram's very recent TED talk about computation.
http://blog.wolfram.com/2010/04/27/step ... ur-future/

This is a very recent article that i highly recommend. It deals with a comparison between the "operating system" of biological entity E.coli and linux. They ask questions of reliability and show their results:
http://www.pnas.org/content/early/2010/04/28/0914771107.full.pdf+html

List of some people who work on parallel computation / software / hardware.
http://openmp.org/wp/about-openmp/

-----------------------

Interesting gesture based 3-d computer OS
http://oblong.com/