Researchers in the field of artificial intelligence seek to produce a smart machine capable of creativity and art making and writing literature, and despite the fulfillment of some of their dreams that machines can write short texts through the development of self-learning techniques, many people still doubts about the ability of machines - whatever their intelligence - to produce music Real or writing exquisite literary texts.

In his report published by the French magazine Le Pen, author Guillaume Galeret quotes the explanations of the philosophy entrepreneur, Thomas Solenac, about his reservations about "deep learning", a technology that is very popular today and relies on a lot of data. This technique has been in popularity thanks to Frenchman Jan Le Can and Canadians Joshua Pingu and Geoffrey Hinton for the past forty years.

Deep learning

This technology is based on the development of artificial neural networks that simulate the way the human brain works, that is, it is able to experiment, learn and develop itself by itself without human intervention, and the "deep learning" technique has proven its ability to recognize images and understand speech and translation from one language to another.

"It is about the structures of artificial neural networks that learn to represent data on a hierarchical basis. Machines also learn to represent the world with multiple levels of abstraction," Yan Le Kun, a researcher at New York University, told the French magazine LeBuan.

To do this, these machines consume, quickly and without absorption, a large amount of data produced by discriminatory analysis algorithms, and deep learning, which is sometimes interesting and boring at other times, depends on a large number of examples whose characteristics must be determined manually, which can be Frequent, and therefore very long. "

However, there is no single way to teach machines to know, as Thomas Solenac explains, who returned to the 1975 controversy between psychologist, Swiss philosopher Jean Piaget and American thinker Noam Chomsky about language learning.

For a specialist in Piaget's epistemology, the common initial instinct for language does not exist, while American linguist Chomsky sees that there are a large number of cognitive structures common to all individuals.

Solinac explains that "the first approach distinguishes deep learning, which considers that a machine is able to learn everything from scratch by 'consuming' a large number of information and data quickly and without assimilation, while the second approach relies on a vision of the world based on the quality of learning more than its quantity" .

  • In his field of scientific specialization, Chomsky is described as the "father of modern linguistics" and the author of the theory of "generative grammar" which is the most important contribution to the field of linguistic theories of the twentieth century. Since joining the institute, he has repeatedly modified his linguistic theory, while maintaining its fundamental postulates. He is also the creator of the Chomsky Sequence theory of linguistic analysis.

Understand the world

The writer pointed out that the second approach, inspired by Thomas Solenac, allows machines to understand the world in a more exemplary fashion.

"Let's take, for example, a company that uses artificial intelligence to quickly respond to emails. When the environment changes dramatically, as was the case with the outbreak of the pandemic, the second approach allows programs to be more interactive," Solenac said. The entrepreneur proposes a technique that does not require training, and already knows the work of the human language in an innate way.

Thomas Solenac explains that this "permits, for example, to confuse and distinguish the word orange, which means orange and orange fruit, which means that the machine may become somewhat understandable."

Culture and carving terminology

However, machines cannot formulate new terms that express new concepts and processes that did not exist before. Philosophers, writers, and even technologists use different creative ways to derive and generate the words they need to express new ideas or techniques. On the other hand, machines lack this creative approach, which requires high analytical capabilities, even for technical concepts.

And machines cannot convey the aesthetic of text and spirit in great literary works, as the expressive ingenuity of writers is a special advantage enjoyed by writers who are able to invent beautiful metaphors to evoke strong feelings, often as a result of hard creative work. Even if machines were able to replace translators in certain areas, literary translation would still be an exclusive privilege for accomplished translators and not others, and in return machines could actually replace "translators" who translate in an "automated" and non-creative way.

As machines lack an understanding of the different cultures around the world, they often fail to recognize the intricacies of special cultural expressions and cannot transfer them to their equivalent in the target language. The context plays a confusing role for machines, while translators serve humans and help them master the transfer of meaning.