[Top] [All Lists]

Re: [ontolog-forum] Semantics of Natural Languages

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Tue, 30 Aug 2011 23:50:27 -0400
Message-id: <4E5DAF83.8040208@xxxxxxxxxxx>
Adrian,    (01)

I agree with your comments:    (02)

> I think John McCarthy once said words to the effect that what's needed
> for the really big problems in AI (such as general computational NL
> understanding), is to stop and _think_ for a very long time.    (03)

Good advice.  But every once and a while, it's useful to document
your thinking, otherwise none of that thought ever gets recorded.
But it's also important to recognize that no single publication
is ever definitive.  Each document is just one step along the way.    (04)

> It's very tempting to hope that something in the philosophical or AI
> literature, combined with the data structure du jour, plus the current
> fashion for AI-style ontologies can somehow be glued together to 'solve'
> general free NL understanding.    (05)

I agree.  Cyc is probably the ultimate development of the classical AI
paradigm of a big ontology + knowledge base + inference engine(s).
About a dozen years ago, I was talking with Mike Genesereth, who said
"Lenat is probably the only one who doesn't know that Cyc has failed."    (06)

I wouldn't say that Cyc is a total failure.  I think that it has some
useful components.  But it is never going to achieve its original goals
if it continues along its current trajectory.    (07)

> However, the evidence points the other way: Watson being beaten by
> Congressman Holt at Jeopardy [1] is only the latest in a painfully long
> list. (Jeopardy competence is much less than full NL understanding anyway)    (08)

Jeopardy is less tightly structured than Cyc, and it therefore has a
better chance to evolve into something useful. But with more thinking,
it could have achieved much better results with much less hardware.    (09)

> So, my money is on someone unexpected, like Einstein in his patent
> office days, with lots of time to think, and not overly wedded to
> current paradigms.  Not rushing to implement.  As they used to say at
> IBM -- _Think_.    (010)

I agree.  The following slides, which I presented at ICCS 2011 in July,
survey 5 different cognitive architectures, of which Cyc is the first:    (011)

    Cognitive architectures for conceptual structures    (012)

Each of the architectures is a moving target that has been evolving
for 20 years or more.  The fifth is the CS architecture, which I
presented in my 1984 book, _Conceptual Structures_.  Those slides
discuss the original 1984 version, the developments since then,
and the relationships to neuroscience.    (013)

John    (014)

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    (015)

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