Robotics Reference
In-Depth Information
Internet access, even though they may not have any training in com-
puter science or Artificial Intelligence, can participate in the Open Mind
database project. If everyone with access to the Internet contributed just
one (different) piece of knowledge, the task would be well on the way to
being completed. So Singh and his colleagues built a Web site located
at
http://www.openmind.org/commonsense/
,
and in its first 15 months
of life more than 8,000 volunteers contributed many hundreds of thou-
sands of pieces of knowledge. This was still only a quarter of what was in
Cyc at that time, but the knowledge in Open Mind was acquired in less
than one-tenth of the time and at a tiny fraction of the cost.
Because the Open Mind project relies on the general public, knowl-
edge is input in plain English sentences, rather than in a precise but
difficult-to-use language
17
such as is employed for the Cyc project. But
the manner in which Singh's project uses plain English has only been fea-
sible in recent years, made possible by the advances in the techniques for
Natural Language Processing, which are now at a level good enough to
translate a large proportion of the sentences people have supplied to the
system into the syntax understood by the software.
Case Based Reasoning
Possibly the main disadvantage of any form of reasoning based purely on
logic, is that the process of searching for a solution to a problem does not
benefit significantly from any kind of knowledge. Instead the process
relies on hopefully intelligent heuristics guiding the creation of the next
step in the solution process, and growing a tree of possibilities until a
solution to the original problem is found.
In the early 1980s, Roger Schank at Yale University investigated the
role that a memory of previously occurring situations can play in problem
solving. This is how we often think when we are faced with a problem.
A Chess grandmaster will often encounter a position on the board that
is similar to positions with which he is familiar, and he knows from his
past experience, his memory of past positions, what moves are the most
promising ones. An airline pilot encountering particularly bad turbu-
lence in an area will often know the geography and weather patterns of
the area well enough to be able to decide correctly in which direction he
should head or to which height. In almost every domain of problem solv-
17
Difficult for anyone who is not familiar with the CycL language.
