Jack Park



Date: Fri, 31 Mar 2000 16:35:23 -0800



Mr. Rod Welch The Welch Company 440 Davis Court #1602 San Francisco, CA 94111 2496
..
Subject:   DKR Design



Miscellaneous comments on POIMS:

plan, perform, report


...is similar to the Lisp programming language's interpreter: read, eval, print and to the inner interpreter of my programs (The Scholar's Companion): find something to do, do it, study what you did.

Communication plays a central role in the POIMS discussion. Several years ago, it occured to me that if we practiced reductionist thinking, we might realize that communication is likely one of just a few really primitive concepts in the universe. My original thinking was that communication started at the protein interaction level (molecular biology), and migrated all the way up to whole cultures. I now think communication starts at the quantum level.
..
One of the other primitives I have deduced is that of the arms race, FWIW.

TheBrain...

http://www.thebrain.com
..
...provides a graphical way to provide Topic Maps...

http://www.infoloom.com

...of intellectual property.
..
LifeStreams...

http://lifestreams.com

...provides a graphical way to record and maintain a time series of event records. The two need to be combined; a project I started a while back to do so is called Amanuensis. It's original incarnation was Java, but it was always intended to become code inside the core of the Linux operating system. I suspect there are profound similarities between Amanuensis and POIMS.
..
POIMS, however, articulates one aspect of the time series issue that I had not given any consideration to in that thinking: planning. Planning is a big artificial intelligence problem and much work has gone into The Scholar's Companion code such that TSC could plan study methods when new/interesting informational events occured. You have brought it out as a first class citizen in the knowledge space. I think that is important.
..
You are steering an interesting line along the border of causal and relational thinking. Newtonian mechanics has taught us to think causally, and only now we are beginning to think relationally. You appear to be plowing a path that brings the two together. Qualitative Process Theory, the overriding paradigm I implemented in TSC tries to do this as well. The mathematics of relational thinking may turn out to be Category Theory, one of the topological algebras.
..
What will make this melded universe work, as we both agree, is a satisfactory response to the ontological issues inherent in this work. In fact, it would seem to me that there is the need to address the knowledge representation issue at the same time. For, without a uniform, though heterogenous by nature, representation scheme, one will find it very difficult to satisfy the demands users will put on the system as they attempt to open or create knowledge spaces against a larger knowledge library (Doug's DKR).
..
It turns out that Oracle 8i, the database, makes an attempt to satisfy part of this problem. As the literature explains it, 8i stores documents in a uniform way, no matter the source. It returns documents in a format driven by the needs of the requesting client. Thus, a Word document can be accessed by a browser in html, for instance.

I suspect that this kind of thinking will necessarily have to go into the final DKR, no matter how it is conceived. In fact, it would seem that some cell phone of the future will also record the call so that it can become part of the DKR. I imagine we will need some new laws about recording conversations for that to happen.
..
My work with Adam is centered around those issues related to knowledge management, and is focussed at this time on those specific issues related to a framework, especially the representational issues. These start with a serialization scheme (likely XML in some form), a representation scheme (several candidates there, including DocBook, some dialect of RDF, and maybe even OML/CKML).
..
There are two issues at work here, especially as relates to your POIMS notions: one is the representation of narrative, and the other is representation of the linkages (relations).
..
Humans tend to grasp the relevance of things drawn in some graphics view. In fact, it turns out that asking kids to draw concept maps (akin to mind maps) is a great way to probe the depths of their understanding.
..
Humans can compose thoughts both graphically and in narrative form. I tend to prefer both: part in narrative -- the background stuff -- and part graphically, taxonomic stuff and process envisionments. Because of that, I can write simple process rules that watch a scenario as it evolves and fire rules to continue the evolution. Thus, I get an envisionment of all the various directions some initial scenario can take. It turns out that if I were able to map a temporal sequence of my own thoughts and experiences to such a graph, I would be able to analyze it in terms of the kinds of rules necessary to fire to bring about the sequence. From that, I can begin to discover mistakes in my thinking, and plan for the future by writing new rules to get it right the next time by firing those rules during a planning phase.
..
With TSC, I wrote a simple constrained natural language interface that allows a user to type simple sentences that get translated into the internal knowledge representation scheme. From there, the user can construct simple queries that play back "what the system learned." I can then use sentences to construct initial conditions and fire rules that build a new envisionment. I hired a high school freshman one year to use it. She wrote a story about the phyto and zoo plankton universe. Her envisionment shows who feeds on whom, and so forth, and a taxonomy was graphically displayed. All from simple sentences. That is precisely where I am personally going with The Scholar's Companion. My hunch is that POIMS is going the same direction, and my desire would be to see the OHS/DKR go that direction as well.
..
Sincerely,


Jack


Jack Park


..
Post Script

this could be restructured into a dialog between us on unrev list.


// Copy to:
    ..


  1. //