Ron and Pat, (01)
Governments and large corporations often have design competitions,
which sometimes produce very good products. The only difference
between a design competition and a free market competition is
whether the choice of winner is made by a committee or by the
customers. Over the long run, I'd trust customers who have full
and fair access to all the information. Unfortunately, in the
short run, the winner tends to be the one with the best sales team,
not the best engineering team. (02)
RW> If we could identify the little groups that are going to produce
> the key ontologies, it would be easy to say "put them in charge
> of ontology now and save all the waste involved in an evolutionary
> survival of the fittest process". (03)
The total amount of money spent on Cyc was over $70 million, mostly
from the U.S. gov't, but some from large corporations. I think that
the money might have been divided up more effectively in a series
of design competitions. (04)
PC> I have no idea what you think is impossible for arithmetic,
> but lots of people have been using arithmetic for very practical
> applications for millennia... (05)
Yes. For simple functions that can be defined in terms of the
"foundational ontology" of +, -, *, and /, people use closed-form
definitions. But for most functions, closed-form definitions are
not possible or practical. They require implicit definitions in
terms of axioms, not closed-form definitions. If you use axioms,
there is no need for any "foundational" terms. That is what I
meant by saying that the ideal number of foundational terms is
either 0 or infinity. (06)
PC> and if we can get a foundation ontology that works as well
> for database interoperability we will be very far ahead of
> where we are now. (07)
Nothing will work as well as arithmetic other than another
simple mathematical system. And it's not at all clear how
"far ahead" a foundational ontology will get us. Remember
that Cyc with its 2 million axioms has not been able to pay
for itself by supporting applications to database design,
knowledge base design, or any other profitable endeavor. (08)
PC> The magnitude of the interoperability problem is so large
> that the government should be funding *every* plausible project
> to solve it. (09)
Perhaps. But the gov't funded Cyc for the past 24 years. It
seemed to be plausible, but it hasn't provided any ROI. Many
other proposals don't feel or smell much different from Cyc.
Why would the gov't funders consider them plausible? (010)
RW> What if a Chinese group gets more industry support than the
> American team; should the US government cut the funding to the
> US group. (011)
Funny you should say that. In 1980, the Japanese funded a major
leap forward called "The Fifth Generation". That was a time when
Japanese cars and other technology were beginning to take over
the U.S. markets. And the Japanese were talking about making
great new progress with artificial intelligence hardware and
software. That stimulated a great deal of investment in AI
projects in the US and Europe -- Cyc, among others, was one of
the US projects. (012)
In 1987, I participated in a Japan-US AI Symposium that was
hosted by the Japanese 5th Generation project. In short, we
felt sorry for them. They had built some nice hardware that
was good by 1980 standards, but the microprocessor technology
had progressed so rapidly that the Japanese special-purpose
hardware was much more expensive and not much faster than you
could get by just putting together a few much cheaper chips.
But there were some good talks at the symposium, and I got
a nice tie-tack that says "5th Generation". (013)
RW> I guess that I have more faith in "socialized medicine"
> than "socialized ontology". (014)
Me too. Every industrialized country except the US has socialized
medicine, and their people are healthier. But socialized ontology,
so far at least, has not made anybody richer, happier, or healthier. (015)
RW> Would $1,000,000 get a complete ontology for the petrochemical
> industry? A small fraction of what they have spent on lobbying
> for drilling in the US continental shelf. (016)
You can get quite far with a much simpler ontology. See the oil
and gas exploration example in slides 26 to 37 of the following: (017)
http://www.jfsowa.com/talks/pursuing.pdf (018)
Two guys from EGI (Energy and Geoscience Institute at the University
of Utah) visited us at VivoMind on a Monday. We ran their 79
documents through our system with our plain vanilla ontology, but
the answers to the queries were not well focused. So we produced
a list of all the terms that weren't in our dictionary (such as
'turbidic mudstones'), and I showed them how to group them in
rough categories, such as Rock, RockFormation, GeologicalStratum,
TimePeriod, etc. On Friday, we ran the same 79 documents and
queries, and the results were spectacularly better. They spent
a couple more weeks to produce the ontology that was used in the
screenshots on those slides. (019)
I believe that we can get a lot out of ontology, but I don't believe
that we should repeat what Cyc did over the last 24 years. It's
important to study Cyc in order to know what was good, what was
mediocre, and what could be done better in some other way. (020)
John Sowa (021)
_________________________________________________________________
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 (022)
|