C ha pte r 1 · O ve rvie w of A rtific ia l Inte llige nc e ■ 25

Scripts 28 are an extension of frames that further exploit the ex-

There are a number of examples
pectancy inherent in human interactions. Through their work, Schank
of early research contributions in
and Abelson have been able to build a number of systems that seem
AI that have been “absorbed” by
to comprehend descriptions of well-defined settings. Conceptual
mainstream computer science; for
graphs by Sowa 29 and Novak and Gowin 30 are a simplified but per-
example, the programming language
vasive heuristic technology that has often been used to represent the
Smalltalk, hierarchical methods, and
knowledge in a discipline.
frames led to advances in the object-
In 1985, Marvin Minsky published Society of Mind.31 Minsky pro-
oriented paradigm.
poses theories to explain the organization of the human mind. He sug-
gests that the intelligent world could be run by agents. These agents
are themselves unintelligent, but could be combined in sophisticated ways to form a society, which
seems to exhibit intelligent behavior. Concepts such as multiple hierarchies, scales, learning, re-
membering, sensing similarities, emotion, and frames are presented by Minsky through the agent
model.

1.5.9 Uncertainty Reasoning

Traditional mathematics often deals with certitudes. The set A is either a subset of the set B, or
it is not. AI systems, much like life itself, are plagued with uncertainty. Chance is an inimitable
component of our existence; for example, you could catch a cold during your morning commute if
the passengers next to you on the bus or the train were coughing or sneezing; then again you might
not. Consider the following sets: the set of people that are satisfied with their jobs, and the set of
people that are not satisfied. It is not at all unusual for some people to belong to both sets. Some
people might love their jobs, though they might believe they are underpaid. You can consider the
set of people who are satisfied with their jobs as a fuzzy set 32 because it varies depending on the
conditions. Often one is satisfied with the work itself. However, one considers the paycheck to be
insufficient. That is, one is satisfied with one’s job to a certain extent. The degree of membership
for a particular person in this set could range from 1.0 for someone who thoroughly loves their
work to 0.0 for an individual who should seriously consider a career change.
Fuzzy sets arise in many domains. Cameras vary shutter speed based upon the amount of sun-
light present. Washing machines control wash cycles that are based, in part, on the dirtiness of the
clothes they contain. Thermostats regulate room temperatures by making sure that temperatures
actually fall within an accepted range, rather than at a precise value. Modern automobiles adjust
brake pressure in accordance with weather conditions. Fuzzy-logic controllers can be found in each
of these devices. A more comprehensive discussion on the role that uncertainty plays in AI is found
in Chapter 8, “Uncertainty in AI.”

1.6 EARLY HISTORY OF AI

Building an intelligent machine has long been a dream of mankind, dating back to the ancient
Egyptians who took “short cuts” by building statues that contained hidden priests who attempted
to provide sage counsel to citizens. This type of hoax has unfortunately appeared throughout the
history of AI; the field has been understandably tainted by such chicanery while trying to achieve
the status of an accepted scientific discipline—artificial intelligentsia.
The strongest foundations for AI stem from the logical premises established by Aristotle (circa
350 BC). Aristotle established the models for scientific thinking and disciplined thought that have
26 ■ Pa r t 1 · I ntro d u c tio n

become the standard of the present day scientific method. His distinction between matter and form
was a forerunner to one of the most important concepts in computer science today: data abstrac-
tion, which distinguishes methods (forms) from the shells in which they are encapsulated, or in
distinguishing the form (shape) of a concept from its actual representation. (Recall the discussion
of abstraction in the sidebar in Section 1.1.)
Aristotle also emphasized the ability of people to reason, maintaining that this facility distin-
guishes us from all other living beings. Any attempt to build an artificially intelligent machine re-
quires this ability to reason. This is why the work of the nineteenth century British logician, George
Boole, was so important; his established system for expressing logical relationships later became
known as Boolean algebra.
Raymond Llull, the thirteenth century Spanish hermit and scholar, was probably the first per-
son to attempt to mechanize human thought processes. His work predates Boole by more than five
centuries. Llull, a devout Christian, set about developing a system based on logic whose aim was to
prove that the tenets of Christianity were true. In his Ars Magna (Great Art), Llull used geometrical
diagrams and primitive logical devices to realize this goal. Llull’s corpus of work inspired later
pioneers, including Wilhelm Leibniz (1646–1716). Leibniz, a great mathematician and philosopher

H uman I nterest N otes purely symbolic treatment of these premises,

sound logical inferences could be made.
G eorge B oole
In the second part of the Laws of Thought,
For a computer Boole attempted to discover a general method
program to exhibit any for translating antecedent probabilities for a
kind of intelligence system of events, to determine the consequent
it was decided early probability of any other event that was
on that it needed to logically connected with the given events.
be able to reason. The
British mathematician, The algebraic language (or notation)
George Boole (1815– that he developed allowed for variables to
1864) developed have interactions (or relationships) based on
a mathematical only two states, true and false. He developed
framework for representing the laws of Boolean Algebra, as it is now known, wherein
human logic. His work consisted of some there are three logical operators: and, or, and
50 individual papers. His main achievement not. The combination of Boolean Algebra and
was the well-known Treatise on Differential the rules of logic enables us to prove things
Equations that appeared in 1859. This was “automatically.” Hence, a machine that can do
followed, in 1860, by a Treatise on the this is in some sense capable of reasoning. 33
Calculus of Finite Differences. The latter was An example of Boolean logic is shown
designed to serve as a sequel to the former here:
work. Boole’s Laws of Thought provided a
general method of symbolic inference, which IF A ≥ B and B ≥ C, THEN A ≥ C
is perhaps his greatest achievement. If you That is the law of transitivity—IF A
are given a logical proposition involving any implies B and B implies C, THEN A
number of terms, Boole showed how, with implies C.