ARTIFICIAL INTELLIGENCE

The artificial intelligence ( AI ) 

AI Response

The artificial intelligence ( AI ) is "the set of theories and techniques used to achieve machines able to simulate intelligence" 1 . It therefore corresponds to a set of concepts and technologies more than to an autonomous discipline constituted 2 . Others, noting the unclear definition of AI, including the CNIL , introduce this topic as "the great myth of our time" 3 .

AI combinations

Often classified in the cognitive sciences group , she uses computational neurobiology (particularly neural networks ), mathematical logic (part of mathematics and philosophy) and computer science . She is looking for problem-solving methods with high logic or algorithmic complexity . By extension it designates, in everyday language, devices imitating or replacing man in certain implementations of his cognitive functions 4 .

Its aims and development have always aroused many interpretations, fantasies and worries expressed in both science fiction narratives and films as well as in philosophical essays .

Definition

The term "artificial intelligence", created by John McCarthy , is often abbreviated as "AI" (or "AI" for Artificial Intelligence ). It is defined by one of its creators, Marvin Lee Minsky , as "the construction of computer programs that perform tasks that are, for the moment, performed more satisfactorily by human beings because they require high-level mental processes such as: perceptual learning, memory organization, and critical reasoning " a , 5. There is therefore the "artificial" side achieved by the use of computers or elaborate electronic processes and the "intelligence" side associated with its goal of imitating behavior . This imitation can be done in reasoning, for example in games or the practice of mathematics , in the understanding of natural languages, in perception: visual (interpretation of images and scenes), auditory (understanding of spoken language) or by other sensors, in the control of a robot in an unknown or hostile environment.

While they are broadly consistent with Minsky's definition, there are a number of different definitions of AI that vary in two fundamental respects  : 6

definitions that link AI to a human aspect of intelligence, and those that link it to an ideal model of intelligence, not necessarily human, called rationality  ;
definitions that insist that AI is intended to have all the appearances of intelligence (human or rational), and those that insist that the inner workings of the AI ​​system must also resemble that of the human being and be at least as rational .

History

Historically, the idea of ​​artificial intelligence seems to emerge in the 1950s when Alan Turing wonders if a machine can "  think  ". In an article "Computing Machinery and Intelligence" ( Mind ,October 1950) 7 Turing explores this problem and proposes an experiment (now called Turing test ) aimed at finding out from when a machine would become "  conscious  ". He then develops this idea in several forums, in the conference "The intelligence of the machine, a heretic idea  " 8 , in the conference he gives to the BBC 3 rd  programMay 15, 1951"Can digital calculators think? " 9 or discussion with MHA Newman, Sir Geoffrey Jefferson and RB Braithwaite 14 andJanuary 23, 1952on the theme "Can computers think? " 10 .

Another probable source is Warren Weaver's 1949 publication of a memorandum on automatic machine translation 11 which suggests that a machine can do a task that is typically human intelligence.

The development of computer technologies ( computing power ) then leads to several advances:

in the 1980s, Machine Learning (Machine Learning) developed. The computer begins to deduce "rules to follow" just by analyzing data;
at the same time, "learning" algorithms are created that prefigure future neural networks, reinforcement learning, support vector machines, etc.). This allows for example in May 1997 the Deep Blue computer to beat Garry Kasparov at the chess game;
artificial intelligence becomes a field of international research, marked by a conference at Dartmouth College in the summer of 1956 12 attended by those who will mark the discipline;

AI Applications

since the 1980s, research has mainly been conducted in the United States, notably at Stanford University under the leadership of John McCarthy , at MIT under Marvin Minsky, at Carnegie Mellon University under Allen Newell, and Herbert Simon and at the University of Edinburgh under Donald Michie , in Europe and China. In France, one of the pioneers is Jacques Pitrat  ;
in the 2000s, the Web2.0 , the big data and new computing powers and infrastructures, allow some computers to explore unprecedented masses of data; it is the deep learning ( "  deep learning  ").
The limits of this area vary and optimize a route was considered a problem of artificial intelligence in the 1950s , and today is no longer seen as a simple problem of algorithm 13 .

Around 2015, the artificial intelligence sector seeks to meet three challenges: the perception of the environment, understanding of a situation and decision making by IA 14 . Producing and organizing massive and quality data, that is to say correlated, complete, qualified (sourced, dated, georeferenced ...), historical is another issue. And deductive ability and relevant generalization of a computer, from little data or a small number of events, is another goal more distant 14 .

Types in AI

Between 2010 and 2017, investments would have increased tenfold, exceeding by 5 billion euros in 2017.ies and techniques used to achieve machines able to simulate intelligence" 1 . It therefore corresponds to a set of concepts and technologies more than to an autonomous discipline constituted 2 . Others, noting the unclear definition of AI, including the CNIL , introduce this topic as "the great myth of our time" 3 .