September 17, 2004
Background material on why a National Project is required
(Communication from John Sowa)
(Footnotes by Prueitt and a reply follows John’s note)
Paul,
My criticism of Cyc was not the focus of the talk, and as I thought I
said very clearly, OWL does *nothing* to solve any of the problems of Cyc [1].
As I have said many times, OWL is 30-year-old technology. The original version of Cyc, which was
started in 1984, began with a frame-like system that included all the
functionality of OWL. Over the years,
they added more and more to Cyc until they have a very rich system of knowledge
representation. They added those
functions for a very simple reason: OWL
was insufficient to do the tasks they started with -- which are essentially the
same tasks that the ARDA challenge is addressing [2].
As I pointed out in those slides, OWL is somewhere between Aristotle's
syllogisms, which are over 2000 years old, and Cyc, which began with an OWL-like
system and kept extending it as the weaknesses and limitations became apparent.
I have nothing against using OWL where it is appropriate, and the Cyc
project includes the same OWL functionality that it started with in 1984. But as Cyc has demonstrated, OWL is not
sufficient. Adding more functionality
to OWL such as RuleML is what the Cyc project did in the late 1980s, and it was
still insufficient. Their project kept
growing for another 15 years, and they still haven't met the challenge of the
knowledge soup.
Topic maps are a very nice notation, which I have recommended as an
informal step along the way to a more complete specification. But they are
similar, in many respects, to a simplified version of some of the diagrams of
UML. I have nothing against UML, which
I also recommend where it is appropriate.
But neither UML nor OWL nor Topic Maps will meet the challenge of the
knowledge soup [3].
That challenge, I believe, is really the central one of the ARDA
project. Please reread the second half
of those slides, which discusses it further.
Then go to the bottom of the slides and reread the references I include,
most of which are available on the web [4]
After you have studied all of the above, then perhaps we can talk seriously
about what to do .
John
John (Sowa)
I do not know if you understand what the BCNGroup has proposed. (see current draft: { *** } )
Frankly, the difficulties related to developing a new approach are mind numbing, and we need for you to understand more about the planning process. Otherwise will we cross paths in the night.
Do you wish to collaborate on core goals related to a demonstration of anticipation from real time harvesting of the social discourse?
The ARDA Challenge Problem workshops will allow some new things to be tested.
Core to this is HIP, which allows and supports the human to make judgments about function. There is a great deal of details about how one might make use of human judgments in the context of Q-SAR and quasi axiomatic theory... all of which I discuss over and over again from various viewpoints.
Paul (Prueitt)
[1] All I was saying was that OWL is where the community is at, and that there are some things to work with in OWL. You may not agree that OWL has any value, but this is not a politically proper place to be. It is arrogant. My hope is that we can show that there is value in what has been done, but that there are also core problems related to the understanding of abstractions and category theory.
[2] We agree.
[3] I disagree. Topic Maps is envisioned by Steven Newcomb and Michel Biezunski depend on human’s making a reification of topics and how topics are represented. As such they are of a different category of knowledge representation than are OWL, Cyc or RDF.
[4] I agree that here is some work that is highly relevant to the development of agile knowledge representation that function more like human knowledge within the mind. It is just that I have not been able to bring a synthesis of how you envision analogy computing. This means that I have not done enough work.