I. 1.2. Historical overview of artificial intelligence
Here are the third significant steps in the history of the
Artificial Intelligence:
· Gestation of the AI (1943-1955):
During this period was carried out the first work which can be
considered as the beginning of the artificial intelligence (even if the term
did not exist yet). We can quote the works of McCulloch and Pitts which was
introduced in 1943 a model of artificial neurons. Some years ago,
Hebb proposed a rule to modify neuron connections. Minsky and
Edmo built the first network of neurons. It is during this period that Turing
published its famous article in which he introduces the test of Turing.
· Expert Systems (1969-1979): the first
expert system, known by the name of «Dendral» was created in 1969 in
order to determine the molecular structure of a molecule being given its
Formula and the results of its mass spectrometer. DENDRAL, like all the expert
systems, is based on a great number of heuristic rules worked out by human
experts. After the success of DENDRAL, other systems of experts were created,
in particular the system MYCIN, which carried out a diagnosis of the blood
infections. With 450 rules, MYCIN succeeded to diagnose at a level close to the
human experts and considerably better than the young doctors.
· The IA in Industry (1980- until now):
At the beginning of the Eighties, the company «DEC» started to use an
expert system for the assistance at the configuration of computer system which
enabled them to save tens of million dollars each year. And then, most of
companies started to be interested in the Artificial intelligence and to train
their own teams of research. The United States and the Japan financed large
projects in AI, and the United Kingdom started again its program of financing.
· The return of the neurons networks (1986-until
now) To the medium of the years 80, four groups of researchers have
discovered the apprenticeship rule "Back-propagation" which allowed the
development of neurons network able to learn the very complex functions
(curiously, this rule dad been proposed in 1969, but had any echo in the
scientific community). Since then year, the automatic apprenticeship has become
one of the most active fields of the AI, and has been applied successfully to
many problems (example the excavation of data).
· The Modern AI (1987-until now): As
time goes by, the artificial intelligence has become more and more a rigorous
and formal scientific matter. The majority of the approaches studied today are
based on mathematical theories or experimental studies rather than on the
intuition, and are applied more often to problems stem from the real world.
I. 1.3. Some fields of artificial intelligence
§ Expert systems: it is in this field
that we perform diagnosis, supervision, or breakdown service of industrial
facility tasks. In our days, medical diagnosis programs are capable to perform
diagnoses which are as reliable as human specialists.
§ Robotics and MAC (manufacture assisted by
computer): This discipline aims at realizing physical agents (robots)
which can act in the world. These robots acquire information with the help of
sensors or cameras.
§ Treatment of the natural language and automatic
translation: This field aims targets the comprehension, the
translation, or the production of the language (written or spoken).
§ Logistics: A program of logistic
planning was used by the American army to coordinate its vehicles, its
equipment and its soldiers during the period of the gulf war in 1991. Thanks to
this system, the problems, which in the past would have taken weeks to be
solved it were solved in a few hours.
| 
