Java Reference
In-Depth Information
were disappointing and did little to bring artificial intelligence applications to
market.
One reason that artificial intelligence programs have not performed as well as it
was hoped seems to be that they simply don't know as much as humans do. In
the early 1990s, Douglas Lenat and his colleagues decided to do something
about it and initiated the CYC project (from enCYClopedia), an effort to codify
the implicit assumptions that underlie human speech and writing. The team
members started out analyzing news articles and asked themselves what
unmentioned facts are necessary to actually understand the sentences. For
example, consider the sentence ȓLast fall she enrolled in Michigan State.ȓ The
reader automatically realizes that ȓfallȓ is not related to falling down in this
context, but refers to the season. While there is a State of Michigan, here
Michigan State denotes the university. A priori, a computer program has none of
this knowledge. The goal of the CYC project was to extract and store the
requisite factsȌthat is, (1) people enroll in universities; (2) Michigan is a state;
(3) a state X is likely to have a university named X State University, often
abbreviated as X State; (4) most people enroll in a university in the fall. In 1995,
the project had codified about 100,000 common-sense concepts and about a
million facts relating them. Even this massive amount of data has not proven
sufficient for useful applications.
Successful artificial intelligence programs, such as chess-playing programs, do
not actually imitate human thinking. They are just very fast in exploring many
scenarios and have been tuned to recognize those cases that do not warrant
further investigation. Neural networks are interesting exceptions: coarse
simulations of the neuron cells in animal and human brains. Suitably
interconnected cells appear to be able to ȓlearnȓ. For example, if a network of
cells is presented with letter shapes, it can be trained to identify them. After a
lengthy training period, the network can recognize letters, even if they are
slanted, distorted, or smudged.
211
212
When artificial intelligence programs are successful, they can raise serious
ethical issues. There are now programs that can scan rÈsumÈs, select those that
look promising, and show only those to a human for further analysis. How would
you feel if you knew that your rÈsumÈ had been rejected by a computer, perhaps
on a technicality, and that you never had a chance to be interviewed? When
