QUESTION 1:
Select any three (3) definitions of Artificial Intelligence and critique each of them in the
context of the four (4) world model. (3 pages max)
The Four World Model: is a view that distinguishes four kinds of world in existence and they
are:
i.
ii.
iii.
iv.
World 0: this is the nucleus world upon which other worlds rest
World 1: the natural world i.e. not man-made
World 2: the created world
World 3: the virtual world.
This model claims that humans exist in the natural world while created things like computers,
houses, furniture etc. exist in the created world. The things made by members of world 2
exist in the virtual world (World 3). This model also claims that members of a world cannot
create things that are also members of that world. This implies that as humans, we cannot
create things that are members of our world.
Hence, if this is so, AI can only mimic the intelligence of members of the natural world i.e. it
can only be like, it cannot be the same as. The four world model will be used as a criteria to
critique three AI definitions in this assignment.
3
2
1
0
4. The goal of work in artificial intelligence is to build machines that perform tasks
normally requiring human intelligence. (Nilsson, Nils J. (1971), Problem Solving
Methods in Artificial Intelligence (New York: McGrawHill))
5. Artificial Intelligence is the use of special programming techniques known as
heuristics to understand and instruct the process of problem solving (Daniel et al
(1986))
6. The branch of computer science that is concerned with the automation of
intelligent behaviour (Luger and Stubblefield, 1993).
Now the critique;
1. The field of Artificial Intelligence (AI) aims to reproduce human intelligence by
artificial means (John N. Hooker (1990) A Postmodern critique of Artificial
Intelligence).
These definition of Artificial Intelligence tells the purpose or aim of the study which is to
reproduce the intelligence of humans (which are a member of the World 0- the natural world)
by artificial means. This definition limits the field of AI to the intelligence of humans. A
question to ask will be if the intelligence of a lizard who is a member of the natural world is
reproduced (Can it be referred to as AI?); this definition says no. I will say yes because
according to the context of the four-world view if a member of the natural world (humans)
are assumed to be intelligent then all natural entities in that world are intelligent.
However, this definition does not tell us by what means the purpose of the study will be
achieved. The artificial means can be generalised to mean any phenomenon not natural. It is
not specific about which artificial means to be used i.e. the kind of study AI is. The field of
Artificial Intelligence as we know it now uses the computational study of the problem solving
abilities of members of the natural word to reproduce their intelligence.
This definition states the object of study to be only humans which is wrong according to the
context of the four-world model. The four-world model assumes all members of the natural
world to be capable of intelligence because, one member of the natural world has to be
assumed intelligent to even know what intelligent means and how to recognize it.
As explained above, this definition can be said to be inadequate because of the reasons stated
above. Therefore, this definition will be termed incomplete and can be rephrased thus:
AI is a computational study that aims to reproduce natural intelligence (their problem
solving capability) by way of artificial means using computational tools.
2.
This definition tells us the kind of study AI is i.e. the use of heuristics in the process of
problem solving which is an important part of AI. This definition is not specific enough about
what kind of problems are to be solved in the field of AI. This definition does not have any
relation to the four-world model as it does not relate this process of problem solving to any
natural entity nor does it relate programming to intelligence.
This definition is lacking because it does not tell us the object of study i.e. who is being
studied and the purpose of the study. The object of study in AI is any member of the natural
world and the purpose of study is to mimic their problem solving abilities. The object of a
study is very important in any field of study as this gives a bound on what can or cannot be
studied
The aim of AI is to mimic or reproduce the problem solving abilities of members of the
natural world through the use of computation which includes but is not limited to heuristics
as perceived from the above definition. Other aspects of computation will come into play
during the reproduction of human intelligence so heuristics cannot be the only technique used
in AI even though it may be the major technique used.
As explained above this definition is not a full definition of Artificial Intelligence and may be
rephrased thus:
AI is the use of computation such as special programming techniques known as
heuristics to understand and instruct the process solving ability of the natural world members
in order to mimic or reproduce it.
QUESTION 2:
Discuss the Turing test of intelligence and provide at least two (2) limitations for why it is not
a sufficient test for human intelligence.
The Turing test of intelligence is ascribed to Alan Turing who wrote a paper titled Computer
Machinery and Intelligence in 1950. The paper was silently retitled Can a machine think?
some years later by a publisher. Alan Turing was a British mathematician and computer
scientist who proposed that instead of asking if machines could think, the question could be
rephrased to if machines can exhibit intelligent behaviour such as cognitive abilities and
decision making.
The Turing test consists of three participants, two competing entities and an interrogator
(judge). He proposed a first phase of the imitation game where the two competing entities
were a man and a woman who were trying to convince the judge (a human) that they were
either i.e. the man was the woman and the woman was the man. The test was supposed to be
performed only through teletype where the test could not be based on any other factor apart
from their cognitive abilities. The interrogator was to ask the other two participants questions
in which their answers was to be used to judge and make decisions about their gender.
The second phase of the imitation game was to replace one of the competing entities either
the man or woman with a machine (computer) and ask the interrogator to ask questions and
guess by the answers to the questions who was human and who was a machine. The ability of
the machine to imitate a human in answering the questions and not its speed or accuracy was
the objective of the Turing test. This implies that the machine could deliberately introduce
some errors in calculations in order to imitate mans level of accuracy and also the machine
could pause before answering question in order to imitate mans speed during thinking. After
the conversation through teletype, the interrogator has to decide which of the competing
entity was human or machine. If the machine could fool the interrogator i.e. deceive the
interrogator into thinking it was human or if the interrogator could not determine who was
human or machine, the machine can then be said to be intelligent.
Machine
Human
Interrogator (Judge)
Fig 2. A Diagram Illustrating the Phase 2 of the Turing Test of Human Intelligence
2. Writing and reading (Literacy) cannot be said to be a key criteria for intelligence
Literacy cannot be used as a key criteria for intelligence as it is a learned skill and not a
natural skill. If humans were taught to read and write, then computers can also be
programmed to read and write and that should not be a criteria to evaluate intelligence.
Intelligence is an innate ability of human beings and therefore cannot be evaluated by a
learned skill.
The Turing test assumes that literacy is a key criteria for intelligence as intelligence is a
cognitive ability that may be sharpened by education whether formal or informal and can be
exhibited by illiterate human beings. The fact that an entity whether human or machine is
literate doesnt make it intelligent rather its ability and capability for problem solving,
capacity for values and judgement, capacity for thinking and reasoning etc. is what makes it
to be considered intelligent. This explains why some illiterate human beings especially our
fore-fathers in Africa have demonstrated high levels of intelligence through their use of
proverbs, riddles and common-sense attitude towards life and other issues in general even
without stepping inside the four walls of an educational institution.
QUESTION 3:
Discuss any three (3) dilemma in AI and state why their definitive resolution is central to the
works in AI.
A dilemma is an issue or situation that presents difficult predicament for the cause of action
or flow of thought (argument) to follow in advancing a precise solution to a problem.
The dilemmas to be discussed are as follows:
1. Whole-Part Dilemma
This is the dilemma of an system A comprising of three entities x, y, z which are functional
parts of the system A such that they define A. Isolating the components x, y or z for study is
not the same as studying these components in A because the components interact with each
other when in A. This means you cannot describe a system in isolation of its components yet
you cannot describe any component alone without it being in the system.
You cannot describe the part of a whole from a whole is not the same as describing the part
of a whole in a whole but is similar i.e. it is an approximation and is known as the
Circumscription Approach.
This can be illustrated using the pie model and the piece in pie model
Also no matter how deep and accurate the study of a human being can be, it can never
capture the entire human person and if it cannot then it should not be termed complete.
The definitive resolution in this dilemma is that a behavioural study should be well defined
from the problem definition of the study. If not the study will be too detailed and become
trivial and not useful.