On 5/31/2011 10:07 AM, Adrian Walker wrote:
> You are right, but unfortunately what you say appears to undermine much
> of the promise of current work on ontologies -- the ability to usefully
> define terms based on primitives that people can read and understand. (01)
There are many, many issues. And it's important to sort them out. (02)
1. A single universal ontology of everything that that spans the full
range from the simplest primitives to the most complex concepts
in every field -- science, engineering, medicine, business, law,
ethics, and everyday life. (03)
Some people talk about this, but there is zero evidence that it's
possible or necessary. Even if it were possible, it couldn't be
achieved until all fundamental questions in every branch of science
are completed. (04)
2. A universal ontology about the "mid range" concepts that everybody
uses in ordinary, everyday life. (05)
Many people have talked about this as a simpler goal than #1 that
could be achieved without solving everything. But if you look
at the history of science, technology, and culture, there has
never been a clear separation of the issues. The cutting edge
science of one generation becomes everyday life for the next. (06)
3. Partitioning ontologies into different domains for different
fields of interest, different business areas, or different
companies. (07)
Unfortunately, every area is interconnected with every other
area. Engineers, lawyers, or finance people from different
businesses or kinds of businesses have more in common with
each other than they have with other departments in their
same company. Nobody has ever found a clean partitioning. (08)
4. Flexible methods for relating independently developed technologies. (09)
There are so many different methods that no simple characterization
is possible. But this is the way that computer systems have been
interoperating for the past 50 years. It is based on incremental
fixes and adaptations, some of which have failed and some of which
have made brilliant breakthroughs. (010)
There are interesting ideas in all these approaches. The one that
has proved to work is #4. (011)
It is conceivable that methods based on some version of #1, #2, or #3
might do better than #4. But they all attempt a "Great Leap Forward".
The problem with great leaps is that when they fail, they tend to
fail spectacularly. (012)
AW
> And of course Watson, which appears to herald the renewed respectability
> "messy" experimental AI. (013)
I would compare Watson to the Semantic Web. The SW attempted to make
a great leap forward by pouring a lot of money into a unified set of
standards that had not been tested in the marketplace. Watson didn't
use SW technology -- it used AI technology from the 1990s, which is
better than the tools that the SW offers today. (014)
Note the recent announcement for the "OCAS Workshop and Challenge": (015)
OCAS
> The real challenge for Semantic Web technologies and ontologies lays
> in the adoption; although the need for this disruptive technology
> is clear, it has not yet been fully adopted by the mainstream. (016)
The SW was first announced in 1994, the foundations (Unicode, XML,
and RDF) were in place by 1998, and SW had a huge international
hype machine with multi-millions of euros behind it. But "it has
not yet been fully adopted by the mainstream." (017)
Compare that to PHP, which was developed by one guy for his
"Personal Home Page" in 1995. It was an incremental improvement
that solved a problem that many people experienced. It took off
like a rocket without any funding from anybody. (018)
For another example, consider Google, which was started by a couple
of graduate students in 1996 and was incorporated in 1998 -- the same
year that the SW base technology was put together. Google was just
an incremental improvement over AltaVista. But today, Google has
far surpassed anything that has been achieved by the SW -- and they
did it without using any SW technology. (019)
John (020)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (021)
|