There are various
approaches to the construction of the systems of an artificial intelligence.
This division is not historical, when one opinion gradually replaces the other,
and various approaches exist now. Besides as there aren’t any really complete
systems of the artificial intelligence at present time, it is impossible to
tell, that one approach is right, and the other is fallacious. To begin with
let’s briefly consider the logic approach. Why has it arisen? In fact the
person is engaged not only in logic insinuations. This statement is certainly
right, but especially the ability to logic thinking distinguishes the person
from animals. The Buleva’s algebra serves as a basis for the given logic
approach. And each programmer is familiar with it and with the logic operators
since he mastered the IF operator. The Buleva’s algebra has received its
further development in the form of calculation of predicates - in which it is
expanded due to the insertion of
subject symbols, relations between them, quantifiers of existence and
generality. Practically each system of
the artificial intelligence, constructed on a logic principle, represents the
machine of the proof of theorems. Thus the initial data are kept in a database
in the form of axioms, or a rule of logic conclusion as the relations between
them. Besides each such machine has the block of generation of the purpose, and
the system of a conclusion tries to prove the given purpose as the theorem. If
the purpose is proved, then the trace of the applied rules allows receiving a
chain of the actions, necessary for the realization of the assigned purpose.
The capacity of such system is defined by opportunities of the generator of the
purposes and the machine of the theorems proving. Of course, we can say that
the expressiveness of the algebra of statements will not be enough for the
valid realization of the artificial intelligence, but it is necessary to
recollect, that the basis of all the existing computers is a bit - a cell of
memory which can accept the values only of 0 and 1. Thus, it would be logical
to assume, that everything what is possible to realize on the COMPUTER, would
be possible to realize in the form of logic of predicates. Though it is nothing
told about the time duration. And such comparatively new direction, as an
indistinct logic allows achieving more expressiveness to the logic approach.
The basic difference of which is that the truthfulness of the statement can
accept not only yes or no (1/0), but also the intermediate values – don’t know.
The great labor
intensiveness is a characteristic for the majority of the logic methods,
because during the search of the proof it is possible to examine all the
variants. Therefore the given approach demands the effective realization of the
computing process, and a good work is usually guaranteed at a rather small size
of a database. As a structural approach
we mean the attempts of construction of an artificial intelligence by modelling
of the structure of a human brain. One of the first such attempts was the
perceptron of Frank Rozenblat. The basic model-based structural unit in
perceptrons (as in the most of the other variants of the brain modelling) is
neuron. Later there were also the other models, which in common people are
usually called as the term "neural networks". These models differ
according to the structure of the separate neurons, the topology of
communications between them and the algorithms of the training. Neural networks
are mostly successfully applied in problems of recognition of images.
The evolutionary
approach has received wide enough distribution too. While the construction of
the systems of the artificial intelligence according to the given approach the
basic attention is paid to the construction of an initial model, and the rules
according to which it can change (evolve). And the model can be made on the
various methods; it can be both neural networks and a set of the logic rules
and any other model. After that we switch on the computer and on the basis of
model check it selects the best of them, on the basis of which by the various
rules the new models generate and
after that the selection of the best are repeated again etc.
One more widely
used approach of the construction of the systems of the artificial intelligence
is imitating approach. The given approach is classical for cybernetics with its
base concept “a black box" (BB). BB is a device, a program module or a
data set, the information about the internal structure and the content of which
are completely absent, but the specifications of input and output data are
known. The object, whose behavior is imitated, is just that very "black
box". It is not important for us, what it and its model contain and how it
functions; the main thing is to make our model to behave the same way in the
similar situations. Thus, the other human quality is being modelled here; this
is the ability to copy what the others do, not going into details, what it is
necessary for. Often this ability saves much time for him.
According to the
philosophical ideas of many scientists in the field of the artificial
intelligence, the consciousness represents rather small superstructure over our
sub consciousness, which watches over the activity of some centers of our
brain, such as the center of speech, the center of final processing of the
visuals, and after that "returns" these objects on the first stages
of the processing of the given information. So
there is a repeated processing of these objects after that, we as though see
and hear, what our brain thinks. Thus while our "active"
participation in the given process
the opportunity of mental modelling of the surrounding reality appears. And
exactly our process of observation over the activity of these few centers is
that we call consciousness. If we "see" and "hear" our
thoughts, we are conscious, if not we are unconscious. If we could model the
work of these few "conscious" nervous centers (the work of which is
based on the activity of rest brain) as one BB, and the work of
"supervisor" as another BB, it would be possible to say with
certainty, that " yes, the given
model thinks and thinks just like me ".
And finishing the
acquaintance with various methods and approaches to the construction of the
systems of the artificial intelligence, we would like to mention, that there is
no distinct border between them. And very often we meet mixed systems, where
the part of the work is carried out on one type, and the other part on another.
Literature
1. Sovetov B.Y., Yakovlev
S.A. The system simulation. Practical course. - Ì.: Vyshaya Shcola. 1999. - 224 p.
2. Waterman D. The
handbook on expert systems. - Ì.: Mir. 1989. - 388
p.
3. Bratko I. The language programming. The prologue to the artificial intelligence. - Ì.:Mir. 1990. - 560p.
4. Robinson J. The logic
programming - the past, the present and the future. - In the book - The logic
programming. Collected articles / Trans. from English and French/ edited by
V.N.Agafonova. - Ì.: Mir, 1988. -
8-26p.
5. A.A.Zinoviev « Higher
education and logic of intelligence », Vestnik Vysshei Shcoly ¹5, 2004. - p.13
6. Korshunov U.M. The
mathematical basis of cybernetics. Ì.: Energoatomizdat.
1987. - 437 – 476p.