Hi John -- Thanks for your clear summary of some of the early history of AI and Databases. You wrote... In the early 1980s, rule-based expert systems were very hot,
and I was working with the AI groups at IBM. Many commercial
applications (including many within IBM) could have benefited
from such technology. But very little of the major AI software
could run on IBM hardware, and none of it could support mainstream
data processing systems that used RDBs and COBOL.
Actually, by the mid 1980s, IBM Yortown Research had the Syllog* prototype -- a highly declarative AI rule system integrated with RDBs, running on IBM mainframes. However, it was neither pure AI nor pure RDB work, and it was ignored by both camps.
It's tempting to see the same situation repeating itself -- with the Semantic Web work now occupying the AI slot. Once again, there's a cultural divide. To experience this, try talking to SW folks about RDBs and SQL, or to database administrators about rule systems and the Semantic Web.
So, as you hinted, the difficulty is as much human-and-cultural as scientific-or- technical. Perhaps this is partly what the push towards "Web Science" is designed to overcome? Cheers, -- Adrian
* See for example http://www.complang.tuwien.ac.at/ulrich/syllog.html http://www.springerlink.com/content/t0h48th74l6u9535/
Internet Business Logic A Wiki and SOA Endpoint for Executable Open Vocabulary English over SQL and RDF Online at www.reengineeringllc.com Shared use is free
Adrian Walker Reengineering On Sat, Aug 30, 2008 at 2:17 AM, Ron Wheeler <rwheeler@xxxxxxxxxxxxxxxxxxxxx> wrote:
John F. Sowa wrote:
> Ed,
>
> Thanks for sending a copy of Chris Welty's note to Ontolog forum.
> Following is a slightly edited version of my response to Chris.
>
> John
> _________________________________________________________________
>
> Chris,
>
> I realize that many intelligent people have devoted a great deal
> of effort to producing the RIF document, but I am disappointed
> by the fact that the Semantic Webbers have isolated themselves
> from the successes achieved by other very intelligent people in
> computer science and data processing over the past 50 years.
>
> For example, the Semantic Web suffers from same provincialism
> that prevented AI from being integrated into the mainstream of
> data processing. To a certain extent, I can't blame the AI crowd,
> because the circumstances that led to their isolation were not
> entirely their fault. I would put some of the blame on IBM, the
> company where I worked for 30 years:
>
> 1. Much of the earliest work on AI in the 1950s was done on IBM
> computers, such as the 704 vacuum-tube machine and the upward
> compatible transistorized versions, the 7090 and 7094.
>
> 2. In 1964, two major developments occurred: IBM introduced
> the new System/360 line, which broke compatibility with the
> 7094, and DEC introduced its first large system, the PDP 6,
> which was as fast as a System/360 Model 65, but as cheap as
> a Model 50.
>
> 3. At that time, both the MIT and Stanford AI groups were
> planning to upgrade their hardware, and they agreed to
> purchase identical hardware configurations so that they
> could share code. The obvious choice was the PDP 6.
>
> 4. The IBM monopoly had sucked almost all data processing onto
> IBM equipment, and DEC was left with a much smaller market.
> But the cutting edge work in AI was carried out on DEC
> equipment until the 1980s, when much of it moved to Sun
> workstations and LISP machines.
>
> As a result of this history, the database field (which used the
> largest available machines, mostly IBM mainframes) was totally
> isolated from the AI research (which ran on much smaller machines
> that lacked industrial-strength database software). Furthermore,
> the COBOL language, which most people in comp. sci. treated with
> contempt, was the most widely-used language for data processing.
>
> In the early 1980s, rule-based expert systems were very hot,
> and I was working with the AI groups at IBM. Many commercial
> applications (including many within IBM) could have benefited
> from such technology. But very little of the major AI software
> could run on IBM hardware, and none of it could support mainstream
> data processing systems that used RDBs and COBOL.
>
> There was once a saying that the economy of the free world ran
> on COBOL and relational DBs. Today, COBOL is less important than
> it used to be, but relational DBs are more entrenched than ever.
> Now the *world* economy runs on RDBs.
>
> Almost every major web site includes a large RDB, and the medium
> sized web sites are based on LAMP: Linux, Apache, MySQL, and
> Perl, Python, or PHP. But the developers of RDF(S) and OWL
> ignored that extremely important part of the WWW.
>
> The Semantic Web failed to provide a seamless integration with
> RDBs. (SPARQL is *not* a solution -- it is just one more step
> *away* from integration.) There is an ISO standard for Prolog,
> which is very widely used for major applications. Experian,
> one of the three largest credit bureaus, *bought* Prologia --
> the company founded by Alain Colmerauer, the inventor of Prolog.
> But instead of building on the ISO standard for Prolog, the RIF
> designers developed an incompatible rule language.
>
> If you want a serious proposal on how to make the Semantic Web
> succeed, following is an outline of my recommendations:
>
> 1. Integrate the SW with relational databases. That means
> n-tuples, not triples, as the primary data format.
>
> 2. An upward compatible version of SQL should be supported
> as the query language, but a typed version of the much
> cleaner Datalog language should also be supported.
>
> 3. Integrate the SW development tools with the mainstream
> software development tools, such as the UML diagrams,
> but provide a firm logic foundation based on logic,
> preferably a logic that puts a premium on semantics,
> not syntax. The ISO standard for Common Logic is
> an example.
>
> 4. Design RIF as a syntactic front-end to ISO standard Prolog,
> but also provide a switch that permits either classical
> negation or negation-as-failure (as in Prolog).
>
> 5. Pay serious attention to Google's protocol buffers, which
> they developed as their primary notation for transmitting,
> storing, and processing large volumes of structured data.
>
> There are many more details to be added, but these are the
> fundamental prerequisites for a successful Semantic Web.
>
> The fact that Google, the largest and most successful of the WWW
> businesses, has ignored RDF and OWL is a sign that the SW, on its
> current path, will be limited to a marginal role on the fringes
> of the WWW.
>
> John
>
> PS Re AI history: Stanford and MIT bought identical hardware,
> but as soon as they wrote the first lines of code, their software
> became incompatible: MACLISP from MIT and InterLISP from Stanford.
> That fragmentation was another obstacle to commercial success.
>
add to such a well thought out set of recommendations but I am coming to
the conclusion that many complimentary and contradictory ontologies are
going to be made freely available and unless we get a structure for
allocating namespaces, we are going to impair the ability of application
developers to assemble a set of foundation ontologies and domain
specific ontologies that can be integrated into a usable base for
customization.
The fact that developers and system manager can depend on unique LDAP
OIDs, unique Java package names and unique Internet Domain Names has
made a huge difference in the capabilities of systems and the Internet.
These are managed in a decentralized way or self managed with a high
level of success.
We are already seeing a few useful foundation ontologies being released .
I am concerned that I will have trouble integrating my application
specific ontology with foundation ontologies produced by different
companies or industry groups if namespaces are duplicated between
vendors or open source projects.
Is this something that IANA (http://www.iana.org) should be mandated to
manage?
Ron
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (01)
|