More on Intelligenesis:
February 8, 1999
PROSPECTUS
A Computer Program That Thinks Like a Human (Almost)
By JANET STITES
Silicon Alley in New York is known for many things,
but not as a hotbed for advances in computer
science.
Mention genetic algorithms or artificial intelligence,
common parlance among computer scientists, and eyes may
very likely roll among even the most respected
new-media executives in New York. In fact, an almost
Luddite attitude toward hard technology seems to
prevail in Silicon Alley, where Bikini.com, "the
world's biggest beach party," is an example of the
cutting edge.
On the fringes of Silicon Alley,
though, can occasionally be found
a company like Intelligenesis
Corp., which was founded in early
1997 by Ben Goertzel, a mathematician, and Lisa S.
Pazer, a former Wall Street economist and analyst.
Intelligenesis is trying to change the way people
measure business risks or make buying decisions by
combining, in a software program called Webmind,
elements of the Internet with artificial intelligence
technology or "machine learning," by which the software
can, for example, "remember" who Alan Greenspan is and
why he matters.
The technology is the brainchild of Goertzel, who
received his Ph.D in mathematics at age 22 from Temple
University, then did research at the University of
Western Australia and later taught mathematics,
computer science and psychology at City University of
New York.
"For a long time, I had been designing an architecture
for artificial intelligence, but didn't have the
machine to do it on," said Goertzel, now 32. "When I
started using the World Wide Web, I realized it was a
metaphor for the brain -- that the Web could be a
mind."
Webmind gathers and processes what it perceives as
"human intelligence" data from text documents, like
news articles -- whether on the Internet or corporate
networks -- to help users spot trends or make
predictions. The results, which can be presented as
numbers, a chart or perhaps eventually even a sentence,
are meant to complement rather than replace traditional
mathematical modeling -- to help analysts, in a sense,
to read between the lines.
Ms. Pazer herself left off trying to outguess Greenspan
and quit Wall Street in 1993 to write a novel. But
after coming across Goertzel's research on the Web, she
put aside her book to start Intelligenesis, where she
now serves as president and chief executive.
Here is how she explains Webmind: "If you want to try
to determine whether Greenspan is likely to raise
interest rates, Webmind can query news archives for
Greenspan's comments over the last six months. Because
it understands aspects of natural language, it can give
a qualitative answer by building a graph giving weight
to each of his comments and showing how he reacted in
the past in relation to his comments. You don't search
for documents; you search for answers."
At full throttle, the software is supposed to be able
to answer specific queries that depend on establishing
relationships, like "How much money did the United
States give to Russia in 1998?"
Ms. Pazer contends that Webmind has far-reaching
applications. "The financial market is our proving
ground," she said. "We can measure risk and volatility.
But the technology could be used by anyone trying to
evaluate risk or isolate trends." The list of potential
users ranges from risk managers at insurance companies
to market researchers to political strategists.
But Webmind is still in its infancy. Intelligenesis is
currently testing the software with potential customers
and has a demonstration for qualified users at its Web
site, www.intelligenesis.net. "We have to prove to them
that the technology works," Ms. Pazer said.
If they do that, Goertzel and Ms. Pazer hope to develop
partnerships, supplying Webmind to financial news
services or other online information providers, in a
revenue-sharing arrangement. As the Webmind brand
becomes more established, Ms. Pazer would look for such
partnerships to give way to straight licensing deals.
Though the business viability of Intelligenesis'
product remains to be seen, the company has attracted
the involvement of some influential people.
The vice chairman, for example, is John B. Evans, a
co-founder and chairman of another Alley-based company,
Biztravel.com. Danny Hillis, co-founder of the
supercomputer company Thinking Machines, who is now
vice president of research and development in Walt
Disney Co.'s department of imagineering, recently
joined the Intelligenesis strategic advisory board.
Among others who are watching the progress of Webmind
is Alexander Chislenko, a researcher at the Media Lab
at Massachusetts Institute of Technology who has
followed Goertzel's work for some time, because it
complements some of his own. "I think it's a credible
technology, and the approach is promising," Chislenko
said. He predicted that Webmind would prove valuable in
modeling some parts of the financial market that other
technologies cannot. But Chislenko cautioned, "No
technology can predict the whole market."
Backed solely by private investors so far,
Intelligenesis employs nearly 30 people, with a heavy
mix of programmers trained in the social sciences --
some working from the company's office in New York
city, others commuting by way of the Internet from
Norway, Russia, Brazil and Australia.
Goertzel and Ms. Pazer contend that all the really
significant predictions and decisions that people make
are, in the end, based on human psychology. And
Webmind, they say, can use artificial intelligence to
translate the multitudinous outpourings of human
psychology into valuable data.
"When you think about Webmind, think about being at a
cocktail party," Ms. Pazer said. "Think if you could
process not just the conversations you had, but all the
conversations in the room."
-------------------------------------------------------
Copyright 1999 The New York Times Company |
