Copy Link
Add to Bookmark
Report
IRList Digest Volume 4 Number 35
IRList Digest Tuesday, 7 May 1988 Volume 4 : Issue 35
Today's Topics:
Query - Metamorph from EPI
Announcment - Metamorph from EPI
News addresses are
Internet or CSNET: fox@vtopus.cs.vt.edu
BITNET: foxea@vtvax3.bitnet
----------------------------------------------------------------------
Date: 12 May 88 10:59:00 PDT
From: "SEF::ROSEMAN" <roseman%sef.decnet@nwc.arpa>
Subject: IRList Digest
...
I also have a question concerning a commercial product called Metamorph from
EPI(Thunderstone/Expansion Programs International Inc.) I read about in an
article passed to me with claims that as a text retrieval and correlation
system this software is using some secret algorithms and is orders of
magnitude better than anything else. I attended the MIT conference in Boston
this spring and don't recall anything about this product or such a
breakthough. Any information or comments would be helpful.
[Note: You asked. See below - Ed.]
------------------------------
Date: Sat, 28 May 88 12:55 EST
From: <VONDRAK%IUBACS.BITNET@CORNELLC.CCS.CORNELL.EDU>
Subject: interest -- Metamorph from EPI
here's a possible article of interest of IRlist.
[Note: This is rather long. I received a copy, asked the submitter to
"sign" the submission and edit it, and to vouch that it was OK with
the publisher to re-distribute this. I still have no "signature" but
since there was the "distributed by permission" comment and since I
had another query (see earlier in this issue), thought I should go
ahead and send this out anyway. I have not seen any technical
details or demonstrations or studies of this "Breakthrough" and so
have no comments. If anyone has further information or evidence to
support the claims made, please let us know. - Ed.]
Defense Science
April 1988
[distributed by permission]
A Technological Breakthrough
by
Harry Zubkoff
Stephen Aubin
Sometimes technological breakthroughs come from the most unlikely places.
Or, perhaps that is precisely where the most significant forms of scientific
discovery are born - where you least expect it. A small, relatively
unknown company in Cleveland, OH, has developed some strategic computer
technology in the information and analysis field that is a quantum leap
ahead of anything else that exists today - the Defense Advanced Research
Projects Agency (DARPA) believes that such technology is many years away.
The technology has wide applications in the defense field, but at present
is being used in limited ways by the US Air Force, NASA, the US Army, the
US Department of Energy, and the Joint Chiefs of Staff (as well as a number
of private sector locations).
To give just a few examples of its wider potential, it could be used by the
intelligence agencies for research and analysis, the Army for tactical
intelligence in the field on its new laptop computers, the Pentagon's
competitive strategies office for use with its simulations of conflict with
the Warsaw Pact, and the National Security Council's Crisis Management
Center. It also has more standard uses: the Office of the Secretary of
the Air Force, for example may use it to more efficiently organize its
voluminous procurement regulations.
The story of this small company illustrates some of the problems with our
procurement system. It provides a microcosm of the scientific discovery and
advances being made across this country by not only well-funded research
efforts in large organizations, but also small- and medium-sized companies
that do not have great resources. But these smaller companies do have some
advantages over their bigger rivals: they are not constrained by the
bureaucracy of larger organizations, for example.
The tale also provides some lessons into how the US must take better
advantage of these advances if we are to remain competitive in the global
marketplace. Once again, it shows in clear definition how the US is
still the spawning ground for scientific discovery, but often fails to take
advantage of these discoveries itself.
Ironically, the technology may stimulate the attention it deserves in the US
through the back door: DoD may legitimately become concerned about such
strategic technology falling into the wrong hands.
Thunderstone/Expansion Programs International Inc. (EPI) has been raising
eyebrows with its text retrieval and correlation products since 1985. Its
recently developed software product, Metamorph, is unlike any other existing
retrieval system. Most people have heard of programs that let you enter a
word or phrase and then find an exact match within some form of indexed text.
Metamorph, on the other hand, can find and match concepts in unindexed text.
Dr. Frederick W. Hegge, a research psychologist for the military, believes
the key to Metamorph's "startlingly powerful capabilities" is its "morphemic"
search capabilities. Morphemes are the smallest meaningful units in
language. EPI, he said, went back to the fundamentals and worked on aning
meaning is carried in the English language. By searching at that level, as
opposed to simply matching words, the way standard retrieval programs do,
"you are one step ahead," he explained.
Metamorph is a software program that incorporates artificial intelligence.
It can correlate concepts by using an intricate search algorithm, the part of
the program that performs extremely rapid string and word comparisons. It uses
this algorithm to look at information via a matrix or web of 250,000
English-language morphemes. Unlike other programs, Metamorph is driven by
natural language. And the operator doesn't even have to use perfect grammar;
Metamorph automatically compensates for those less gifted in the use of the
English language.
Metamorph is also very fast at what it does. It is so fast that one highly
credentialed academic, who works at a high-tech firm and specializes in
artifical intelligence, accused EPI of fabricating the demonstration because
he could not believe the speed with which Metamorph went about its task. In
an IBM/AT microcomputer, Metamorph can search unidexed ASCII (a common
language employed in the most personal computers) text at the rate of
500,000 characters per second.
"We started working in an area of software beyond simple data retrieval,"
says Michael Pincus, CEO of EPI. "We were interested in going one step past
existing software and creating a product able to operationally correlate
information. Our idea was to take much of the correlation burden away from
the human - to the extent it is possible with a computer - so as to free up
human mind to go to the next step, which is a finer correlation of
correlated data."
Pincus pointed out that an article in "Scientific America" last August
discussed how researchers at AT&T Bell Laboratories were "proud of themselves
because they had been able to build a little program that could correlate a
matrix of around 60 words and do it fairly well with a body of text." "We
thought it was funny," says Pincus, "for at the time, we had just succeeded
in building a matrix that had in excess of 100,000 associated words in it.
That, of course, has grown to well in excess of a quarter of a million words
in matrix."
Pincus was surprised in other ways, too. For instance, what his program was
able to accomplish on a microcomputer, he assumed the government had already
developed for mini- or mainframe computers, "simply because the government
spent so much money on the subject of information analysis, content analysis
and so on."
In fact, last July, Dave Ross, who runs a syndicated radio program called
"Chip Talk," interviewed Bob Simpson of DARPA, who was working on machine
intelligence research. [Metamorph had been out on the street for seven
months at this time.] The project was aimed at designing a system of field
computers that could collect and organize battlefield information about
enemy forces and then crank it through the "brains" of the Army's best
strategists, whose expertise would be programmed into the system.
Simpson pointed out during the interview that if such a system had existed
during the incident involving the USS Stark in the Persian Gulf, the captain
might have been alerted to the hostile intent of the Iraqi aircraft. As he
pointed out, the captain knew the Iraqi plane was out there, but had no way
to determine or "interpret" its intent as hostile. Curiously enough, the
kind of system the Captain of the Stark needed already existed.
The first people who ever bought Metamorph was the Japanese. "When the
product was first advertised, I could not keep the Japanese out of our
office," says Pincus. In one instance, a Japanese businessman arrived with a
suitcase containing documents authorizing the transfer of a quarter of a
million dollars in cash to EPI if the company would immediately give him
the rights to a Japanese conversion of the product, along with the exporta-
tion and importation rights to Japan. Pincus declined.
As he explains, "We had to fend off some considerable attention from the
Japanese while at the same time not being stupid and failing to sell the
product ... The Japanese are all over our society looking for opportunities
in high-tech, particularly in artificial intelligence because of their own
fifth-generation artificial intelligence group."
EPI is a company of seven employees. And even though he has to pay attention
to the bottom line, Pincus is leery of allowing the Japanese and other
foreign customers to procure his most sophisticated software programs,
including a number of versions of Metamorph. He describes his company as
pretty "red, white and blue," and believes that what he is doing could
greatly benefit the US government.
In fact, Pincus is very careful about what version he sells to foreign
customers, as well as who the foreign customers are. But, he notes, even
technology legitimately sold to US allies tends to make its way to the
Eastern bloc and the Soviet Union. He already knows, for example, of one
unauthorized copy of a less powerful version of Metamorph that ended up
somewhere in West Germany and then disappeared.
In another instance, a Polish software company proposed a joint venture in
software development with EPI. Inquiries were also received from a
Czechoslovakian firm. Once again, Pincus declined on both counts. And, at a
1987 international conference on artificial intelligence and robotics in
Japan, while EPI's distributor was there demonstrating Metamorph, the only
country that was ready with the cash to make the purchase on the spot was the
Soviet Union. When the distributor called Pincus for approved for the sale,
he denied the request.
Pincus is confident that his company's software cannot be "reverse
engineered," since it was built using previously unknown techniques and
complex algorithms.
There is a certain type of chauvinism that exists in the scientific world.
Oftentimes, if a scientist hasn't been able to replicate the success of
another, he simply dismisses the possibility that what the other claims
actually exists.
A perfect example of this is the now well-known case of Stanford R.
Ovshinsky's "amorphous" circuit technology. Because Ovshinsky was not
connected with any acknowledged research center, his breakthroughs were
disregarded by many US scientists in the early 1960s and 1970s.
Ovshinsky's work, however, was not ignored by the Japanese.
Seeing an opportunity, the Japanese, working through Sharp Electronics,
financed Ovshinsky, who had become utterly destitute and declared bankruptcy.
Today, his co-owned factories are in Detroit, but the US government is forced
to buy amorphous circuits from Sharp.
A similar phenomenon exists in industry and government. Some have dubbed his
NIH, the "not-invented-here syndrome." If a company didn't create the
product, or the US government did not fund its research and development, then
it must not exist or be very useful.
There is also the question of the US technology base and the government's
role in sustaining the United States' technological lead in the world.
Recently, the government created disincentives to what is called independent
research and development (IR&D) through its procurement policies. EPI used
its IR&D funds to create Metamorph. IR&D consists of research projects that
companies themselves initiate, fund, and manage in order to develop new
technologies. IR&D funds are distinct from R&D funds, which are provided by
the government and applied to fairly rigid government specifications.
True innovation, more often than not, comes from the private sector.
Small and large companies alike need the freedom (and financial
security) required if they are to take risks in developing new technology.
Congress, in its wisdom, has placed ceilings on IR&D funds. Many in industry
have pointed out that such disincentives discourage US scientific curiousity
and ingenuity. In a nation where maintaining the technological edge over
potential adversaries is such a high priority, it is odd to see policies
emerge that threaten US industry's ability to continue to do so.
Robert B. Reich, writing in the "Atlantic" a year ago, echoed that point.
"Americans continue to lead the world in scientific discoveries and Nobel
laureates," he wrote. "But we have had difficulty turning our basic
inventions into streams of commercial products. We tend to get bogged down
somewhere between the big breakthrough and its application."
Reich cites a number of examples. American scientists invented the solid
state transitor, for instance. However, in 1953 Western Electric licensed
the technology to Sony. Sony made dramatic improvements and launched
a line of high-tech electronic products. Similarly, he notes, Unimation, an
American company in the forefront of industrial robotic technology in 1963,
licensed Kawasaki Heavy Industries to make industrial robots. The
result: the American robotics industry never got off its feet; the
Japanese, however, have flourished. According to Reich, the same pattern
exists for video-cassette recorders, basic oxygen furnances, microwave
ovens, computerized machine tools, integrated circuits, and so on.
While the US has had its share of technological breakthroughs, this generally
means less to the nation in military or economic terms than the speed and
success with which it is "absorbed improved upon, and incorporated into new
products and processes," writes Reich.
Pincus' new software seems to be another potential area for the Japanese to
exploit, if they can unlock its secret. Unfortunately, American industry and
government have yet to publicly recognize its value as a technological
breakthrough. EPI is, however, deluged on a daily basis with requests for
technical information from all major private and government research groups.
Many seem to want the secret while discounting the usefulness of the
Metamorph product. Again, the not-invented-here syndrome?
There are a number of questions that need to be posed based on the story
about this small Ohio-based company engaged in creating new leading-edge
technology. For one, will that company, and others like it, be there tomorrow?
Reich described in his article how the US forfeited its lead in the produc-
tion of memory chips. In 1980, he writes, US companies were producing most of
the world's memory chips. The Japanese then entered the game. Today, of the
15 American companies who were producing memory chips, only three are still
around and they are all in the red.
Failure to take the next step once a new technology is invented involves
serious economic and military implications. As Reich points out, the US
National Security Agency now buys all its ceramic packages (used to house
chip circuits) from Kyocera, a Japanese company. Other examples abound.
What will it mean for the US if other countries capitalize on American know-
how because the US is unable or willing to apply its own technological
breakthroughs? An even more serious question involves how well the govern-
ment provides incentives to the private sector so that companies of all sizes
will continue to innovate and develop new technlogy.
Initiatives like these are good. However, what the US seems to lack is an
overall vision for the Twenty-first Century. After leading the world in
innovation and technology for over 40 years, will the US become a follower,
forced to play catch up as other countries pass us technologically?
Information is power in today's world. Possessing it, managing it, analyzing
it, and making sense of it before others do, are keys to staying out in front
economically and militarily.
It is almost ironic that nobody wants to believe that a seven-man company
might hold part of the answer to maintaining the United States' competitive
edge in the world that is completely dependent upon communicating and
receiving information.
Mike Pincus continues to marvel at the amount of attention that is company's
new technology has received from technical people and foreigners, but how
little interest many in established companies and in the government show for
a potentially revolutionary product.
Metamorph represents another US breakthrough in technology. Let's hope the
Japanese are not the ones who reap its benefits. If they do,we have only
ourselves to blame.
####
------------------------------
END OF IRList Digest
********************
