Artificial Intelligence - Rick Wagoner

vinegarclothΤεχνίτη Νοημοσύνη και Ρομποτική

17 Ιουλ 2012 (πριν από 6 χρόνια και 6 μέρες)

403 εμφανίσεις

Artificial Intelligence - 1 -

Artificial Intelligence

Rick Wagoner

Information Education and Technology 600, Section 002
Professor Dr. A. Zargari
Artificial Intelligence - 2 -

November 14, 2004
Artificial Intelligence
Artificial Intelligence has been a fantasy in the minds of people for many years.
In the movie ‘A Space Odyssey: 2001’ the computer HAL was a product of this
imagination. However, artificial intelligence or AI as it is referred to by so many is
beginning to move from the realm of fantasy and imagination to the world of reality and
everyday life. AI is making headway even though quite slowly. Part of the reason for
this that there are so many ambiguous terms and definitions that it is truly difficult to
pinpoint when AI will be a true reality.
Many of the questions to be answered deal with the definitions being used to
describe the processes needed to develop AI. In response to this we hope to refine our
understanding of what AI is which in turn requires that we define intelligence and
learning. Once we have defined the important terms and developed a usable
understanding of AI we will look at some applications of AI at this time and also how AI
can either benefit or harm in its use.
When dealing with trying to develop an artificial intelligence we first must define
what intelligence truly is. However, this is not a trivial task. According to (IQTest, 2004)
in 1921 an academic journal asked 14 prominent psychologists and educators to define
intelligence. They received 14 different answers. This experiment was repeated in 1986
with 25 experts. Again the answers widely varied by individual. There was however a
couple common threads in the many different answers. Experts in both experiments
emphasized two abilities in defining intelligence. The ability to learn from experience
and also the ability to adapt to a changing environment where listed as key elements of
Artificial Intelligence - 3 -

intelligence. This does not define intelligence in concrete terms and there truly does not
appear to be a concrete definition of intelligence. (Hearst & Hirsh, 2002) also show this
somewhat ambiguous definition by stating “a chief essence of human intelligence is
continuing learning or adapting, and we rely on that all the time.” One thing that comes
up in these definitions is the ability to learn. However, this also is an extremely
ambiguous term. What does it mean to learn?
Learning is a way to increase knowledge and thus give a better basis for decision
making. (Wills, 2004) defines learning in a slightly different manner than most experts.
He states “an organism is said to have learnt when it has increased its options for
applying new or different behavior to a specific set of circumstances which the organism
believes will be to its benefit.” This definition simply means that and informed deci sion
is a better decision. However, (Wills, 2004) also states that learning is hard to monitor or
track because learning is now always immediately shown. Quite often learning is stored
for use at a later time.
So learning is a way to increase our options and intelligence is the ability to learn.
However this may not be the case. There are still several questions that are being
discussed in the context of learning and intelligence. (IQTest, 2004) asks “is intell igence
one general ability or several independent systems of abilities? Is intelligence a property
of the brain, a characteristic of behavior, or a set of knowledge and skills?” We still do
not have concrete definitions of either learning or intelligence and may not for some time
to come listening to the experts discuss the topic. But for our discussion we will use the
definition as listed above. Learning is a way to increase our options in similar future
situations and intelligence is the ability to learn.
Artificial Intelligence - 4 -

According to the definitions we have developed artificial intelligence is a system
designed with intelligence. This means that the system can learn to possibly change
behavior in the future which is the primary key to intelligence. This elementary definition
goes along with the Turing test which is the primary accepted test to determine whether
AI has been achieved. (Dobrev, 2000) “The Turing test is quite simple. We place
something behind a curtain and it speaks with us. If we can’t make difference between it
and a human being then it will be AI.” This test was developed by Alan Turing who was
a British mathematician who became famous for deciphering German codes during World
War II. Now we have an elementary definition of AI but is this how AI is truly viewed by
the experts?
(Artificial, 2002) “Artificial Intelligence is a branch of science which deals with
helping machines find solutions to complex problems in a more human-like fashion.”
This definition implies that the machine must have intelligence and be able to learn to
solve the problems in a human-like manner. In most cases AI is initially thought to be a
branch of study in the field of computer science. However, upon closer inspection we can
determine that AI covers many different fields of study (Artificial, 2002) including
computer science, math, psychology, biology and philosophy as well as many others.
Intelligence and the ability to learn are not limited to a particular domain. Intelligence
and the ability to learn are not respective of a particular field of study but should be able
to be applied and implemented across the board, regardless of the problem domain being
Why do we pursue an artificial intelligence? There are many reasons for so doing
and there also are benefits to be gained as well as possible harm to be encountered.
Artificial Intelligence - 5 -

(Hearst, 2002) “AI first demonstrated that important intellectual tasks could be
accomplished by selective heuristic search, often in a thoroughly human way.” Many
applications that are currently in use today use technologies made available through AI
research and advancements. The majority of the virus scan engines in use on today’s
desktop computers have the ability to do a heuristic search. This simply means that the
scan engine detects a pattern in the information be searched that could possibly be a virus
but that it also does not have a definite positive match for. The one element that keeps
this from truly becoming AI is that learning is not truly taking place according to our
definition. The virus scan engine is not storing the findings of its search so that its
behavior may be changed in the future. It’s not truly AI but it is a start.
There are many possible future applications of AI. (Artificial, 2002) They range
from military applications to use by the entertainment industry for use in computer games
and robotic pets. With the ongoing situation in Iraq at this time the military applications
really caught my attention. Through the use of AI we could give artificial intelligence to
a robotic pilot of an attach helicopter who could then fly the many dangerous missions.
The intelligence would allow the pilot to correctly target enemy combatants as well as
fully control the helicopter to provide needed air cover for ground troops. This takes an
actual human life out of the equation. Without AI, the robotic pilot is unable to learn
what constitutes an enemy versus an ally. AI gives the pilot the ability to change future
decisions and actions through reference of the current situation. The possible uses of AI
are as limitless as the imagination.
There is also possible harm to be encountered from using AI. Take our military
example from above. The possibility of improper targeting or attack could create a major
Artificial Intelligence - 6 -

catastrophe. As with any new technology there are always ethical considerations for the
implementation and use of the technology. The same situation exists with AI. Given the
current level of AI research and development there is still a good amount of time to
discuss these issues in detail and determine the best path to deal with each.
One thing I came across in my research really made me think. (Brave, 2003)
discussed a study that was completed in the United States. This study determined that
55% of the meaning in human-to-human communication comes from facial cues. We see
as much meaning as we hear when talking with another person. Considering this
information I have to wonder if artificial intelligence will ever be an attainable goal.
Thinking back to Turing’s test and the need to carry on a conversation with an AI artifact
and not be able to know it makes me wonder if this is possible when half of our
understanding from a conversation comes from the facial cues of our conversation
partner. Can we develop and build a system that can fully mimic a human being in both
thought and action? It is something to consider.
Regardless of the concern for attainability at this time, AI is a research topic that
gives light to some very fascinating possibilities. There is one thing about humans that
make us ‘human’. We adapt and change our behavior according to the situation and
environment to ensure continued viability. I found a quote during my research that states
the same thing but the quote is discussing Artificial Intelligence and its use. The quote is
(Hearst, 2002) “Find a bug in a program, and fix it, and the program will work today.
Show the program how to find and fix a bug, and the program will work forever.” This is
the goal of Artificial Intelligence.

Artificial Intelligence - 7 -

Artificial (2002). Artificial Intelligence: Plain and Simple. Retrieved on November 12,
2004 from
Brave, David (2003) AI think, therefore I am. Retrieved on November 12, 2004 from
Dobrev, Dimiter (2000). AI-What is this: A definition of Artificial Intelligence. Retrieved
on November 12, 2004 from
IQTest (2004). Retrieved on November 12, 2004 from
Hearst, Marti A., Hirsh, Haym (2002). AI’s Greatest Trends and Controversies.
Retrieved on November 12, 2004 from
Wills, Mike (2004). Definition of Learning. Retrieved on November 12, 2004 from