“exploring the
history of PLa To
and the wild story
of the people who
designed, built, and
used it will enrich
everyone’s overall
perspective on the
world’s embrace
of the internet and
computers today,”
says Donald Dear.

Illinois.” One such alumnus is Marc Andreessen, who considers himself fortunate to be a product of that environment.

In computing history, the story of
this educational computer system is
little known outside of the circles of
people who either directly used the
system or helped to build it. Invented
by Bitzer in 1960, PLATO, an acro-
nym for Programmed Logic for Auto-
mated Teaching Operation, began as
a project at the University of Illinois
at Urbana-Champaign and was even-
tually commercialized in the 1970s
by Control Data Corporation. Over
the years, the system developed into
a sophisticated, networked comput-
ing infrastructure consisting of smart
terminals that could run games, chat
rooms, and courseware. It was used
for education and training by many
schools and universities, corpora-
tions, and by several branches of
the U.S. government, including the
military.

based in los angeles, Kirk L. Kroeker is a freelance editor and writer specializing in science and technology.

Artificial Intelligence

Computer Scientists Beat U.S. Stock Market

A pair of computer scientists has developed an artificial intelligence program that, in simulated trading, outperforms the standard & poor’s (s&p) 500 Index and six of the top 10 quantitative funds.

developed by Robert p. schumaker, an assistant professor of information systems at Iona college, and hsinchun chen, mcclelland professor of management Information systems and director of the Artificial Intelligence Lab at the university of Arizona, the Arizona Financial text system, or AZFintext, analyzes the text of financial news in the same fashion as a Wall street analyst. however, AZFintext boasts

an enormous computational advantage over a lone human analyst with its ability to scan large volumes of financial news and analyze the words in the text.

current computer-aided quantitative funds analyze numerical data, not text, and schumaker and chen’s program is unique in that it takes a human analyst’s approach of deciphering financial news.

In terms of its ability to
analyze one news article,
AZFintext “won’t be as accurate
as an individual analyst,” chen
told The Wall Street Journal in a
recent interview. “the computer
is maybe 80% to 85% accurate
when analyzing text, but it can
read maybe 100,000 times the
amount of data.”
AZFintext’s technique
involves surveying financial
news and stock prices, and
buying or shorting stocks it
believes will increase by more
than 1% in the next 20 minutes.
AZFintext sells the stocks after
20 minutes. “When you do long-
term predictions, there are many
variables,” chen said. “But … you
can have an advantage if you look
at five minutes, 10 minutes.”
six of the 10 leading quantitative
funds. schumaker and chen also
used quantitative strategies to
produce a stock portfolio and
employed AZFintext to make
trades. With this quantitative
approach, they realized a return
in excess of 20%.
chen expects AZFintext to be
deployable for real-world trading
in two to five years.

schumaker and chen tested AZFintext with five weeks of data from the fall of 2005, which included more than 9,000 news articles and 10 million stock prices. AZFintext produced an 8.5% return on its trades, outperforming the s&p 500 and