"Artificial intelligence: a good servant and a bad master"

Philippe Lecigne is a partner in a prospective consulting firm.

He was editor-in-chief of the reviews Résumé (Groupe Express-L'Expansion) and Belvédère.

He publishes this column within the framework of the Pont-Neuf Foundation.

You can find the entire report “

France against robots?

Automation and Artificial Intelligence

In recent years, the latest developments in computer science, qualified in a somewhat arbitrary manner "

", seem to crown an evolution towards a world where man is confined to the state of spectator, even of useless protagonist.

They revive the fears expressed by Georges Bernanos in 1947 in his essay on France against robots or by Stanley Kubrick (

) twenty years later.

Read also:

Isn't the discourse on Artificial Intelligence overdone?

Is it real intelligence and is it able to surpass human intelligence one day?

To avoid falling into the trap of demonization, or the opposite of unreasonable technological optimism, the first thing to do is to understand what we are talking about, and to distinguish between what today is proven and what is pure speculation.

The term Artificial Intelligence, which dates back to the 1950s, covers very different concepts and issues.

Traditional or symbolist AI - what computer scientists call "

" - is based on a set of rules and leads to clear solutions: from the amount of your monthly income, the price of the property you want to acquire and from your personal contribution, the bank can calculate the amount of your loan, its duration and the amount of repayments that you will have to make.

We can clearly see that “General Artificial Intelligence”, which is the stage where the machine is supposed to overtake man and render him useless, is today a matter of science fiction.

The "

"

approach

based on neural networks, which has undergone dazzling developments for less than ten years, opens up new perspectives and poses quite other problems.

An algorithm is trained to recognize characteristics from a large number of data (for example signs indicative of pathology in data obtained from medical examinations) with a reasonable probability, therefore with a certain risk of error.

The strength of the software lies in its computing power and its memory capacity.

But he is devoid of intuition and common sense.

The computer is a good servant and a bad master.

Real use cases, in industry, medicine or finance make it possible to identify the potential of AI but also its limits.

We can clearly see that “

”, which is the stage where the machine is supposed to overtake man and render him useless, is today a matter of science fiction.

Read also:

The deployment of Artificial Intelligence tools nevertheless raises a number of fundamental questions.

First, there is the risk that machines (robots or software) will replace humans in a number of tasks and lead to increased unemployment.

For the OECD, 10% of jobs would be affected, which would represent a doubling of category A job seekers in France!

The numerous analyzes published in the United States (Acemoglu, Restrepo, Autor), but also in Europe, allow us to answer two questions: does automation systematically involve the destruction of jobs and is it not a the causes of the disappearance of the middle classes in industrialized countries?

We will see that the answers are not identical in Europe and in the United States.

The financialization-automation-offshoring triptych has wreaked havoc on the other side of the Atlantic.

The situation in Europe, where we have more particularly analyzed the economic studies on Germany and France, is more nuanced, but the risk of social polarization is nonetheless a matter of concern.

Let's face it frankly, it is indeed an incompatibility of civilization between Europe and these two countries that their visions of artificial intelligence reveal.

Secondly, questions that arise in terms of public freedoms, risks of bias and discrimination, of explicability of the results of algorithms, which have not yet found fully satisfactory answers.

A detailed analysis of the rather brutal approaches developed in the United States (“

” for example) or in China highlights the limits of the so-called “

” approach of Artificial Intelligence.

Google claims not to do evil when it takes over our data and China considers that the systematic registration of the population through its Social Credit system is only for the good of society.

Let us say it frankly, it is indeed an incompatibility of civilization between Europe and these two countries that their visions of artificial intelligence reveal.

Therefore, the answer can only be political.

Read also:

This is why we have outlined at the conclusion of our report some avenues for strategic reflection for France and its European partners, in an attempt to define an Artificial Intelligence policy for the years to come around two complementary concepts: a normative approach. - term in our sense more appropriate than the vague notion of “

”

AI

- and the search for complementarities between the contributions of AI and the skills specific to humans.

As Montherlant invited us to do in his beautiful text on the Civil War, let us never forget that "

"!