What context?

The premises: the Turing test

Artificial Intelligence (AI), in its current meaning, finds its starting point in the 1950s with the work ofAlan Turing, who wondered if a machine could "think". In this approach, this researcher developed a tool to measure the intelligence of the machine, called the Turing test. In concrete terms, a person communicates in parallel with another person and a machine over a long period of time, without visual or auditory contact, for example via a chat program.

If, after the conversation, the tester cannot tell with certainty which of the interlocutors is human and which is a machine, the machine has passed the test and can be considered intelligent. The American sociologist Hugh G. Loebner awarded a $100,000 prize in 1991 for a computer program that passed the Turing test and thus fooled a panel of experts. In 2015, the film Ex Machina made headlines by proposing a pictorial reading of this test.

 

Increasingly efficient technologies

The research area "Artificial Intelligence" attempts to simulate human perception and action using machines. The increasing development of computer technologies, including computing power and algorithmic techniques (notablydeep-learning ), has made it possible to create computer programs that surpass humans in some of their most emblematic cognitive abilities.

Some computer manufacturers have contributed to the media coverage of the cognitive capabilities of their Artificial Intelligence systems. For example, in the context of marketing operations, like IBM with the six chess games played in the 1990s between the Deep Blue supercomputer and Garry Kasparov, the world champion at the time.

 

AI, what are we talking about?

We can consider different devices intervening, together or separately, in an Artificial Intelligence system such as :

  • automatic reasoning
  • machine learning
  • automatic integration of information from heterogeneous sources

 

A copy of the human?

The partial reproduction of the cognitive capacities of the human being allows the optimization of the numerical data present in the large databases of certain companies. Thus, an algorithm trained for this purpose can learn from the data it manipulates, and draw new information from it.

However, emotions can never be separated from human cognition. Intelligence is above all emotional, communicative, social, physical, and by nature backed by consciousness. What a machine can never have.

 

Limited intelligence

This artificial intelligence, considered by specialists as still being of low level, compared to the performances of the human brain, has nevertheless the advantage of allowing a system to create new information, thus added value, in an automated and continuous way, via algorithms.

The increasing use of these algorithms, by public and private entities, raises fundamental moral and ethical questions. Can we let a machine make a decision that affects a human? Is it possible to understand how an algorithm works when the machine is self-learning? In case of illicit negative impacts caused by an algorithm on humans, which responsibility should be brought into play: that of the creator of the algorithm or that of the legal entity managing the AI?

 

Philosophical and social issues

The Parcoursup case

These major philosophical, and therefore social, issues must be framed to avoid any drift in the public and private use of AI algorithms. The case of Parcoursup is significant of the negative effects that algorithms can have on a large number of humans, in terms of inconsistent decisions, but also of the questionable influence that programmers can have on the legislator.

Indeed, some remarks made by the computer scientists who created the algorithm on which Parcoursup is based have led the legislator to modify the rules of the ORE law concerning the maximum rate of non-residents and the minimum rate of scholarship holders in assignment decisions. Is it acceptable for a legislator to have to revise his copy because of technological imperatives?

 

Bias at the center of the debate

In order to avoid the biases and perverse effects of the current rush to AI, national and European regulations have been adopted over the last four years.

They will be analyzed in a second article to come. A third article will then be devoted to a complete chronology. It will address the issue of AI and the human challenges it raises.