Friday, August 13, 2004
On the separation
of syntax and semantics
And the issue of (real
time) pragmatics
John Sowa issues raise the same issue I raised with Bjorn as far back as 1995.
Part of the issue is that Bjorn (Gruenwald – holder of the Primentia patent on a Hilbert encoding of data structure) understands his invention better than I do, but does not have the background in formal systems that I do. Then again, I am trying to communicate the underlying theory for formative and differential ontology and I am trying to work on the details as to how Sowa’s cognitive graphs might work in a Hilbert Engine environment. MITi/Readware’s substructural ontology will help us refine more of this communication. See [75]
The, to be proposed, project to combine a syntactic procedure, the Hilbert Engine, with a theory of semantics, (the Readware ontology of language), and a theory of pragmatics (formative and differential ontology developed by myself from my reading of semiotics) will draw from the strengthens of several points of view. A conjecture about what might be discovered using this type of technology is given in bead [46] of the “graph representation of knowledge bead thread.
We have gone beyond surface, and have begun the definition of stratified theory, but everyone needs to actually know, for example, Sowa’s work in cognitive graphs:
This is why there should be an objective and "honest broker" type evaluation of the data encoding methods. In my opinion, we have to start with the PriMentia engine [1] because of the differences in speed and memory requirements that are obtained.
Our intent has be to ask DoD think tanks to serve this role, and failing that to go directly to the Congress. This is an on going process that has a number of well-recognized problems with turf and politics.
We have funding proposals in, and have prepared to accept venture capital.
The national need is for deep exposure of what our community of scientists knows about the issues. This cannot happen without disturbing some of the traditional structure that provides advice to the Congress and Administration.
We have some specific problems with business processes, and the things that have been going on for decades but which seem to have become extreme since 9-11-2001.
For example, if one simply looks at the LSI (latent semantic indexing) patent and the $10 - 50 million spent on systems development obtained by the patent owner, SAIC, because of the LSI ownership, then one sees that most of what is being purchased has huge and non-appreciated holes in the foundational theory and use. The patent owner is often neutral on issues of scientific judgment, preferring to let the client identify what the truths are and to build as specified by business contracts, rather than allow a scientific type investigative process governed by a community of scholars.
There are positives and negatives to business processes being in charge. We are open to think tanks and trusted consultants as being the “honest broker”, but they have to recognize that when a group of outside scientists come up with new innovations that they, the think tanks and consulting groups, have some responsibility to pass this information on to the government clients.
The most important “scientific problem” is the problem of formalization of induction and the sometimes-deceptive utility of deductive reasoning. Many scholars, Sowa, Penrose, Kugler, Rosen etc; have addressed these sources of errors. But the utility of deductive reasoning in mathematics and the success of mathematics in engineering have been seductive.
Almost all of the issues that Sowa raised about the Hilbert encoding frame the problem with induction in cognitive neuroscience and behavioral theory:
We propose to extend the foundations of mathematics in ways envisioned by Lev Goldfarb:
and to support a broad investigation of the BCNGroup Conjecture about Periodic Tables.
Goldfarb develops an exposition of how the numerical model is developed in elementary mathematics, and then suggests that the numeric model is not the only type of inducted formalism possible. He is not alone in this suggestion.
Our work on the concept of formative and differential ontology is a simple exposition of how very small ontologies, in the form of cognitive graphs, can be constructed using a pass over, or convolution, the set of pairs of nodes with relationships in the form
{ < a, r, b > }
where the individual < a, r, b > are acquired through some type of data mining process.
Because the convolutions can be underconstrained, the inductive process is left to the last moment and can be made to depend on algorithmic processes such as associative (artificial neural network) memories, evolutionary programming, or human supplied constraints (applied in real time via some type of semiotic interface). All of the details of this type of “knowledge operating system” (KOS) have yet to be worked out, but in principle the KOS allows one to delay the formation of deductive mechanisms and to make this formation process sensitive to the real time interaction from humans.
A knowledge sharing core concept was developed almost two years ago to address this "evaluation of technology" problem:
I lay out the architecture again [66] [70].
In the architecture, the Hilbert encoding is used rather than a hash table, so one simply uses the Hilbert encoding as a data encoding.
I talk about the Hilbert encoding as a “key-less hash table.
My claim is that the new technology can do things that others cannot. It is a simple matter to test this claim. I claim a number of other things that are beyond what has demonstrated. The Knowledge Sharing Core was designed to test claims of this type.
I use convolutions over the points of a discrete Hilbert line. This is the basis for the machine side of the Anticipatory Web. HIP (Human-centric Information Production) then depends on "ontology services" like what Readware has developed, as well as topic map type reification (human checking and comment).
If we go slowly we can sort this out, while at the same time laying down the CCM patent from Applied Technical Systems and the SchemaServer from Schema Logic. These technologies should also be compensated as a National Project is defined and the investigation about the BCNGroup Conjecture about Periodic Tables.
Peter Krieg and Richard Ballard both have something to say at this point, as does Sandy Klausner (CoreTalk).
I suggest that we each go slow, talking about what our contribution is and inquiring carefully as to what the claims are.
I will try to summarize the best parts of the discussion.
John (Sowa) perhaps you would make some comments on the Readware work,
or the ATS patent: which I have generalized to my open source IP at:
The ATS patents used to be linked from the web site, but I do not seem to be able to find them now.
[1] Wednesday, September 01, 2004. Since this bead was posted, the BCNGroup has become aware of prior art to the Primentia patent and has become involved in a project to use this as an alternative to the Primentia “Hilbert Engine”.