- From: Henry Story <henry.story@bblfish.net>
- Date: Sat, 17 Dec 2005 13:04:24 +0100
- To: Joshua Allen <joshuaa@microsoft.com>
- Cc: "Frank Manola" <fmanola@acm.org>, <semantic-web@w3.org>
I think all of this is way too pie in the sky for the semantic web. Yes the Semantic Web is about machines communicating with machines. But that does not mean you have to go the whole way towards machine learning and advanced AI technologies. Oracle is a multi billion dollar company. It produces databases that are uses by other software to both store and extract data. There we have "machine to machine" communication. Everybody uses a database. But most Mom and Pops don't access the data using SQL. They use a machine to automate that process. So with the Semantic Web. The reason you use ontologies, is the same reason you use classes in Object Oriented programming, or you use table structures in SQL databases. They help humans, who write the software that does machine to machine communication understand and master what they are doing. The contracts can then furthermore be used to tell when the software is doing what it should be doing, ie. to tell when there is a bug. Most machine to machine communication I would say is of the sort described above. The AI sort, which admittedly is very interesting, is not what happens on the bulk of the internet. There is a lot less magic in the Semantic Web as some would hope for. But that is for the good. Most of the technology it builds on is very well understood. Henry Story On 17 Dec 2005, at 00:46, Joshua Allen wrote: > >> or not. Re "simply ignore the dictionary and talk to each other", > this >> is a bit simpler if there's a human in the loop than for strictly >> machine-to-machine isn't it? > > I find it interesting that many of the KR researchers have moved to > "machine learning". There is one approach, which says that > communication (human-human or machine-machine) must follow a > "standard" > or "contract". There is another, which says that communication is an > organic process of learning. Much interesting machine-machine > communication these days is done through machine learning rather than > taxonomy/ontology/contract.
Received on Saturday, 17 December 2005 12:04:33 UTC