[Top] [All Lists]

Re: [ontolog-forum] IBM Watson on Jeopardy

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Mon, 14 Feb 2011 11:58:14 -0500
Message-id: <4D595F26.7070503@xxxxxxxxxxx>
Pierre, Krzysztof, Simon, and Doug,    (01)

I'm returning to the original subject line in order to emphasize
the continuity with the issues about Watson.    (02)

>> But a top-down, monolithic, detailed, universal ontology of everything
>> is not only impossible to achieve...    (03)

> The nature of the alleged impossibility is unclear and the various
> characteristics do not seem particularly well on a par as prospective
> causes.    (04)

In notes to the SUO list and Ontolog Forum for the past ten years,
I've been giving reasons upon reasons why it is futile to search
for a perfect top-level ontology.  Yet many people have continued
to propose such things.  Every one of those proposals has been
severely criticized as inadequate, usually for multiple theoretical
and practical reasons.    (05)

> You will excuse the non-native speaker if my command of English is
> not sufficient to see through the apparent paradox that something
> allegedly impossible to achieve may also lead to disaster when acted
> on. So, is the warning here that somebody will lose their mind in the
> raving lunacy of illusory omniscience?
> What, precisely, is said here?    (06)

The disaster can arise from people in a position of power who ignore
the criticisms and edict their choice as a requirement to be imposed
on everybody else.    (07)

The most successful standards are *de facto* standards that people
voluntarily choose because they find them useful.  But the only
widely accepted de facto standards are terminologies and things
like WordNet that nobody considers adequate as a formal ontology.    (08)

The lack of widespread support for universal ontologies is evidence
that none of them are sufficiently mature to be considered as
an official standard.  However, many ontologies are very good and
quite useful as special-purpose modules (or microtheories).    (09)

>> I could not agree more [with JFS], this would be almost like
>> a conceptualization oligarchy. The even more important point
>> however is that it is impossible and we should stop doing it.    (010)

> This is a very pessimistic view on progress, in science, and human
> matters in general. Granting the impossibility of the task for the
> sake of the argument (as noted above, I am not entirely sure of what
> the task at hands is), great things can be achieved when trying the
> impossible.    (011)

I agree that there should never be a prohibition on anybody who
wants to pursue his or her insights as pure research.    (012)

Unfortunately, there are people who attempt (and sometimes succeed)
in converting untested research proposals into official standards.    (013)

> I'm not sure that Cyc is necessarily the best example of a top-down
> and monolithic ontology; a lot of the more detailed areas show clear
> signs of middle out development driven by specific projects. The basic
> idea of  microtheories at makes a start towards enabling modularity;
> it's the presence of the gigantic macrotheories where everything seems
> to depend on everything else that makes it hard to figure out.    (014)

I agree.  But in 1984, when the Cyc project began, the goal was to
develop a monolithic ontology and knowledge base.  After a few years,
they recognized the inevitable conflicts that arise between different
points of view that are equally valid for different kinds of problems.    (015)

During the 1990s, Cyc was completely revised, reorganized, and
modularized into a basic backbone hierarchy plus microtheories.
The current version has over 6000 microtheories.    (016)

In that revision, most of the detailed axioms were removed from
the upper levels and pushed down to the middle and lower levels.    (017)

I have a very high regard for the hundreds of person-years of hard
work that went into the development of Cyc, and I believe that it
is a valuable resource.  But at the same time, I would recommend
a methodology closer to Watson for developing a system designed
for question answering and commonsense reasoning.    (018)

> One often sees that several contestants on Jeopardy are pushing the button,
> but one wins on timing.  How's that going to work out on Jeopardy?    (019)

We'll see tonight at 7 pm EST.  Stay tuned.    (020)

John    (021)

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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (022)

<Prev in Thread] Current Thread [Next in Thread>