Sunday, December 04, 2005
Center of Excellence Proposal
à
[bead thread on curriculum
reform]
The Taos Institute
(on the possibilities)
Communication from John Sowa,
Jeff,
I was going over my email from the past week, and I don't think that our views on the SemWeb and related issues are too far apart. People are often confused because I sometimes attack formalization and I sometimes support it.
My basic point is that human thought and language are capable of logical precision and formalization when it's appropriate and of arbitrary degrees of flexibility when that's appropriate. At the end of this note is a copy of a note I sent to the ONTAC WG about the impossibility of having a fixed, frozen, universal ontology. But at the same time, I also think that a modular, flexible framework is not only possible, but extremely valuable.
Following is a summary of my position:
1. The expert system technology of the 1970s and '80s was not bad for what it did, and there were many successful applications. However, it was not easy to train an average programmer to use that technology effectively.
2. Of all the technologies that came out of that era, logic programming and the Prolog language were the most successful, and they have been widely used for years, especially in Europe. But there were many successful applications in the US as well. Unfortunately, the hype machine in the US put too much emphasis on inferior technology.
3. RDF was designed by Tim Bray working with Guha, who had been associate director of Cyc in the early '90s. But Tim has admitted that the design of RDF was "broken", and he recommended a redesigned version. I agree, but the W3C did not take Tim's advice.
4. There is a need for storing data in a portable format, and if decent tools are available, they could enable a crappy form, such as RDF, to work effectively. (Microsoft, for example, has abundantly demonstrated the power of garbage + hype + money.)
5. If somebody paid me to process a batch of data in RDF and OWL, I would hold my nose, convert it to a better form, do the processing, and convert the output back to the crappy format they asked for.
6. And I believe that some of Lenat's lack of success is his own fault. For years, I was trying to tell him to devote at least 25% of his resources to applications that could (1) bring in some cash that would help pay for the research, and (2) help show what kinds of approaches worked best. But he didn't want to dilute his focus on research. I believe that was a major blunder because applications are necessary to test and guide any kind of research.
John Sowa