In a report voted in plenary session on 14 March, the NATF deems it necessary to build an “AI plan” at European and national level that will make it possible to reinforce investments in computing resources and technological research. The Academy recommends supporting and encouraging a responsible approach to AI that aims to simultaneously develop the performance and ethical acceptability of intelligent algorithms.

Designed to penetrate the majority of practices and environments, in business as in civil society, AI represents a double challenge – competitiveness for our companies and national sovereignty.

AI and automatic learning use a wide range of methods and the first merit of this report, which is particularly addressed to companies, is to explain how they complement and can enrich each other: what is the range of possible solutions? Which methods to choose according to the problem to solve? And, most importantly, what should be the essential steps in an AI strategy in terms of skills, data collection and practices?

The  experts point out that AI is not only defined by scientific methods and algorithms, it is an engineering discipline in which technology and practice play an essential role. However, while there is an excellent French research community on the core domains of AI and automatic learning, France, as a global industrial ecosystem, is lagging behind the United States and China. It becomes necessary to invest in terms of resources and critical mass in computing resources, data volumes and team size to be competitive with the best global players.

The report recommends the development of testing and certification laboratories by business area. Indeed, according to the NATF working group, priority must be given to focusing on new ecosystems that use artificial intelligence methods around business domains, with their own data sets, computing resources and specific business skills.

This report contains recommendations to public authorities, in particular with regard to certification, auditability and restoration of civil society trust. In particular, the Academy recommends the creation of a frame of reference associated with the notion of “responsible AI”, proposed and recognized by the research community. This reference framework could be associated with an artificial intelligence observatory, placed under the control of a European agency, whose objective would be to monitor and reference AI practices and usages in our society. In addition, the Academy recommends a relaxation of the modalities of application of the general regulation on data protection (GDPR) in order not to weaken European ecosystems compared to American competitors.

(Report in French)