Marseille edit
IA, kezaco?
The term "artificial intelligence" was created by the American John McCarthy, considered as the pioneer of the discipline beside Marvin Lee Minsky, Claude Shannon and Nathan Rochester in 1956 in Dartmouth College of New Hampshire in the United States. We can define it as a concept to set up machines feigning the human intelligence. The term is ambivalent because it also indicates a discipline, which studies the intelligence thanks to the tools of data processing and who, in its continuation, allows to create an artificial intelligence in its first direction.
But from which moment can we really talk about intelligence? There is no objective criterion. Is intelligent what seems to be intelligent: everything depends on the context and each subjectivity. The notion is more vague as we know an only single shape of intelligence: human beings. Yet, IA could turn out different. Moreover, it is more just to talk about artificial intelligences in the plural to reflect its multiple realities.
Alan Turing, the inventor?
The mathematician and cryptologue British Alan Turing is considered the inventor of the artificial intelligence (l’intelligence artificielle). During the Second World War, he created the machine Enigma which allows to decipher the communications of the German armies and contributes to the victory of the Allies.
Then, it is at the origin of a test, the imitation games, which raises the question of the difference between the human intelligence and machine. Initially, it is about a fabrication through which a person questions it for one second. According to the written answers (they do not communicate by the voice and do not see each other), it is going to try to know if her interlocutor is a man or a woman. Then, Turing widens the test to IA. In 1950, it predicts that in the edge of the years 2000, the examiner will be less than 70 % of luck to reveal the real identity of his interlocutor during this game in a five-minute duration. In reality, it happened in 2014 when the chat-bot Eugene Goostman deceived ten from thirty examiners during a competition organized in the Royal Society of London by Smelt Shah, a lecturer at the university of Coventry and Kevin Warwick, a professor of cybernetics to the university of Reading.
However, according to the specialist Jean-Gabriel Ganascia, talking about the inventor is maybe excessive because of the inadequacies of the test. Furthermore, it did not create the term. However, its contribution is that we can qualify Turing of "the most influential precursor ".
IA, Queen of games
IA is known of the general public for its victories in various games at the highest level of their discipline. So, in 1997, the Deep Blue IBM computer beat the world champion of chess, Garry Kasparov for the first time. In 2016, it's the turn of Go game to grin and bear IA with the victory of AlphaGo, the program of Google DeepMind company, against the Korean, Lee Sedol, one of the best in the world. To note that this board game of Chinese origin has been considered as out of IA’s reach for a long time, at least for a while, because of its complexity owed to its intuitive character where chess games are led by a mathematical logic.
Except for these two media events, IA has been distinguished in the checkers and the tic-tac-toe in the years 1950-1960. More recently (from 2015 to 2017), it placed influence over human beings upon Jeopardy and Poker. It is even an exploit because, on the contrary to the failures and the Go game, Poker is based on an imperfect information system (all the information is not visible for each of the players)
What’s the deep learning?
We talk a lot about "deep learning" or "deep learning" at the moment. It is a system of learning and classification that works thanks to a network of artificial neurons. It is so powerful that it can make complex tasks as the recognition of the image contents or the spoken language understanding and overtake the human brain with these precise points.
Moreover, the deep learning is inspired by the functioning of the latter. It consists of layers of (digital) neurons which realize each calculation. The aggregation of these operations makes the IA power. We talk about "deep" learning because the greater power of the computer in the past years allows to align numerous layers producing billions of connections.
The father of the deep learning is a French: Yann LeCun. He manages FAIR today, the center of fundamental research on the Facebook IA.
Weak IA, strong IA
We talk about weak IA to indicate that dedicated to the simulation of specific cognitive functions as the image or word recognition. It is against the strong IA which would be the following stage with robots endowed with consciousness. It is advisable to specify that originally the notion of strong IA is a critical concept by the philosopher John Searle towards the cognitive psychologists who considered machines as models of the consciousness and thought.
The concept was resumed by a movement of thought: the trans-humanism. Taken by the American, Ray Kurzweil, the director of Google engineering, they think that IA will exceed that of human beings because of its fast development at a precise moment: that of the Peculiarity dated in 2045.
Science or fiction? For the philosopher, Jean-Gabriel Ganascia, the Peculiarity is a myth which is a profession of faith more than the scientific prospective. Why? Because if they can have a technical autonomy, machines are unable to establish their own laws and give purposes to their behavior.
This debate on the strong IA shakes fears maintained by years of the science fiction (Terminator) cinema. It tends to hide the colossal progress which has already been achieved by IA and concrete problems put into the daily paper as the biases. The IA’s current strength is this combination between the "automobile learning" (allowed in particular by Deep Learning) and the data processing of mass. Yet, these are produced by human beings biased themselves. There is thus a risk that IA also turns to it. In this respect, we remember Tay, the interactive IA of Microsoft which published racist messages on Twitter in 2016.
However, it is advisable not to underestimate the trans-humanism brilliance which counts a number of followers in Silicon Valley. Yet, the latter arrange ways of their ambition within GAFA and other monsters of the new technologies. Moreover, to prevent dangers of a strong IA, they recommend a hybridization of the Man and machine as shown by the project Neuralink of the same name company and behind which we find the boss star Elon Musk.
What is ít idea? Increase our brains thanks to electronic implants, to avoid the enslavement of human beings from IA.
Article written by Thierry Randretsa