Joel, (01)
JB> I am whole heartedly thankful that I have the opportunity to
> pursue being an innovator and early adopter in a business and
> business climate that has become almost entirely consumed by
> establishing a direct and causal relationship between any and
> all activity and "the bottom line". (02)
That is a worthy goal, and I wish you the best of success. (03)
But there have been many "boom and bust" cycles in the field
of artificial intelligence and related areas since the founding
conference in 1956. The booms were caused by overhyping modest
demos to get funding. The busts were caused by the inevitable
failures of projects that could not deliver what they promised. (04)
Examples: (05)
1. Fully automated high quality machine translation (FAHQMT)
was the goal of several well-funded projects in the US,
Europe, and the Soviet Union in the 1950s. In the US,
the DoD funded many prominent researchers (including
Noam Chomsky). (06)
2. In 1964, the ALPAC report said that FAHQMT was beyond the
state of the art, and the added keypunching costs of typing
in the source material would make FAHQMT more costly than
translation by bilingual humans. (07)
3. That report killed almost all funding for MT around the world.
Even the Soviets said that if the Americans didn't believe it
could be done, they weren't going to waste money on it. (08)
4. But the Georgetown Automatic Translator (GAT), for which funding
was terminated in 1964, was a good automated dictionary of words
and phrases. For translating physics articles from Russian to
English, it was not as good as human translators, but it was
faster and cheaper. GAT was commercialized as Systran, which is
still widely used today. A version called Babelfish is available
on the WWW. (09)
5. In the early 1960s, a version of neural networks called
perceptrons were hyped for all kinds of pattern recognition.
They were useful for some purposes, but they were overhyped,
they couldn't meet expectations, and funding for all versions
of neural nets dried up for over 20 years. (010)
6. In 1972, Terry Winograd's SHRDLU system was touted as doing
"natural language understanding." There was a recession,
and people needed funding, so they overhyped it. And they
failed to meet the promises. That caused another bust for
AI developments, especially natural language. (011)
7. In the early 1980s, expert systems looked promising, and
people overhyped the possibilities. They got funding for
the Fifth Generation Project in Japan and competitors in
Europe and the US, such as Cyc. They failed to deliver
on the promises, and successful AI projects changed their
names to "object oriented" to avoid the taint of "AI". (012)
In the past few years, there has been an enormous amount of
hype for ontology and the Semantic Web. Some of the projects
are good, but others have been so wildly overhyped that they
will bring down the good projects with them. (013)
JB> My take is that John is being obstreperous to goad me
> into explaining myself more succinctly... (014)
The word 'obstreperous' is too mild a term. It's more accurate
to say that I was disgusted by yet another half-baked idea that
was thrown into the oven before the dough had a chance to rise. (015)
I thank Chris P. for bringing Kuhn's article to our attention
and Hervé for the URL: (016)
http://content10.wuala.com/contents/nappan/Documents/Kuhn,%20Thomas%20S.%20%281961%29%20-%20The%20Function%20of%20Measurement%20in%20Modern%20Physical%20Science.pdf?token=1261942071525 (017)
As I said, the field of ontology is highly unsettled, there are
some promising ideas, some modest successes, a great deal of hype
about technologies that are still unproven, and some wildly over
funded projects that are almost certainly destined to fail. (018)
Several people have suggested topics that any budding ontologist
should study. I agree with most of those suggestions, many of
which I also mentioned in my books in 1984 and 2000. But as Chris
observed, the field of applied ontology has not yet reached a stage
of maturity for a meaningful Ontology Aptitude Test or certificate. (019)
I believe that *any* OA test we could produce today is likely to
award certificates to people who would be unqualified to solve the
problems of actual paying customers. That would make 'ontology'
a term of ridicule that nobody would put in their resumes. (020)
John Sowa (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)
|