You are on page 1of 4

Chapter fifteen

Concluding remarks
15.1 Benefits
This book has discussed a wide range of intelligent systems techniques and
their applications. Whether any implemented intelligent system displays true
intelligence whatever that is assumed to mean is likely to remain the
subject of debate. Nevertheless, the following practical benefits have stemmed
from the development of intelligent systems techniques.
Reliability and consistency
An intelligent system makes decisions that are consistent with its input data
and its knowledge base (for a knowledge-based system) or numerical
parameters (for a computational intelligence technique). It may, therefore, be
more reliable than a person, particularly where repetitive mundane judgments
have to be made.
Automation
In many applications, such as visual inspection on a production line,
judgmental decision making has to be performed repeatedly. A well-designed
intelligent system ought to be able to deal with the majority of such cases,
while highlighting any that lie beyond the scope of its capabilities. Therefore,
only the most difficult cases, which are normally the most interesting, are
deferred to a person.
Speed
Intelligent systems are designed to automatically make decisions that would
otherwise require human reasoning, judgment, expertise, or common sense.
Any lack of true intelligence is compensated by the systems processing speed.
An intelligent system can make decisions informed by a wealth of data and
information that a person would have insufficient time to assimilate.

2001 by CRC Press LLC

Improved domain understanding


The process of constructing a knowledge-based system requires the decisionmaking criteria to be clearly identified and assessed. This process frequently
leads to a better understanding of the problem being tackled. Similar benefits
can be obtained by investigating the decision-making criteria used by the
computational intelligence techniques.
Knowledge archiving
The knowledge base is a repository for the knowledge of one or more people.
When these people move on to new jobs, some of their expert knowledge is
saved in the knowledge base, which continues to evolve after their departure.

15.2 Implementation
Since intelligent systems are supposed to be flexible and adaptable,
development is usually based upon continuous refinements of an initial
prototype. This is the prototypetestrefine cycle, which applies to both
knowledge-based systems and computational intelligence techniques. The key
stages in the development of a system are:
decide the requirements;
design and implement a prototype;
continuously test and refine the prototype.
It is sometimes suggested that the first prototype and the first few
revisions of a KBS be implemented using an expert system shell that allows
rapid representation of the most important knowledge. When the prototype has
demonstrated its viability, the system can be moved to a more sophisticated
programming environment. This approach makes some sense if a mock-up is
required in order to obtain financial backing for a project, but it also brings
several disadvantages. Working in a shell that lacks flexibility and
representational capabilities is frustrating and can lead to convoluted
programming in order to force the desired behavior. Subsequent rewriting of
the same knowledge in a different style is wasteful of resources. An arguably
better approach is to work from the outset in a flexible programming
environment that provides all the tools that are likely to be needed, or which
allows extra modules to be added as required.
Software engineers, particularly those working on large projects, have
traditionally been skeptical of the prototypetestrefine cycle. Instead, they
have preferred meticulous specification, analysis, and design phases prior to

2001 by CRC Press LLC

implementation and testing. These attitudes have now changed, and rapid
prototyping and iterative development have gained respectability across most
areas of software engineering.

15.3 Trends
Intelligent systems are becoming increasingly distributed in terms of both their
applications and their implementation. While large systems will remain
important, e.g., for commerce and industry, smaller embedded intelligent
systems have also started to appear in the home and workplace. Examples
include washing machines that incorporate knowledge-based control systems,
elevators that use fuzzy logic to decide at which floor to wait for the next
passenger, and personal organizers that use neural networks to learn the
characteristics of their owners handwriting. Communication between
embedded applications is likely to extend further their influence on our daily
lives.
In addition to being distributed in their applications, intelligent systems
are also becoming distributed in their implementation. Chapter 9 discussed the
blackboard architecture for dividing problems into subtasks that can be shared
among specialized modules. In this way, the right software tool can be used for
each job. Similarly, Chapter 5 looked at the increasingly important technique
of intelligent agents. The growth in the use of the Internet is likely to see
increased communication between agents that reside on separate computers,
and mobile agents that can travel over the net in search of information.
Furthermore, Jennings argues that agent-based techniques are appropriate both
for developing large complex systems and for mainstream software
engineering [1].
Paradoxically, there is also a sense in which intelligent systems are
becoming more integrated. Watson and Gardingen describe a sales support
application that has become integrated by use of the World Wide Web, as a
single definitive copy of the software accessible via the web has replaced
distributed copies [2].
As a further aspect of integration, computers are required to assist in
commercial decision making, based upon a wide view of the organization. For
example, production decisions need to take into account and influence design,
marketing, personnel, sales, materials stocks, and product stocks. These
separate, distributed functions are becoming integrated by the need for
communication between them. (Use of computers to support an integrated
approach to manufacturing is termed computer-integrated manufacturing, or
CIM.) However, smaller-scale systems are likely to remain at least as

2001 by CRC Press LLC

important. These include intelligent agents that serve as personal consultants to


advise and inform us, and others that function silently and anonymously while
performing tasks such as data interpretation, monitoring, and control.

References
1.

Jennings, N. R., On agent-based software engineering, Artificial


Intelligence, vol. 117, pp. 277296, 2000.

2.

Watson, I. and Gardingen, D., A distributed case-based reasoning


application for engineering sales support, 16th International Joint
Conference on Artificial Intelligence (IJCAI99), Stockholm, Sweden, vol.
1, pp. 600605, 1999.

2001 by CRC Press LLC

You might also like