Conditions for implementing Artificial Intelligent technologies in critical systems: interests and risks in their application. The Thales’ case

The discipline of Artificial Intelligence is by now present in our everyday lives. The industrial sectors in which this intelligent technology is applied to optimize processes, improve performance, enhance a better customer experience, etc., are a lot, but each of them must be treated and analysed individually. Critical infrastructure systems (Defence, Aerospace, Air and Ground Transportation, Energy sectors and others), namely those in which life is potentially in danger, make an extensive use of Artificial Intelligence, but its application is not as simple as it might seem. This therefore raises questions about what are the necessary conditions for the use of Artificial Intelligence in these sectors, conditions that ensure the minimization of all the errors that could have dramatic consequences and the maximization of the level of reliability of the intelligent models used. Therefore, the present Master Thesis work has as its main objective the research of the essential requirements for the application of intelligent technologies in all the activities carried out in sectors where life is constantly at stake and the analysis of their potentialities and limits. The investigation begins with an in-depth literature review that examines AI, from its origins to the challenges it has faced and the main positive results obtained from its application. Following, the concept of interpretability of Artificial Intelligence is analysed and the key role that it covers in critical systems is explained. In this work, conceived in the form of a case study, Thales Group and some of its projects are deeply inspected. These projects are clear practical applications of Artificial Intelligence technologies, with all their risks and defaults, and confirm the validity and necessity of the conditions researched not only in terms of minimization of terrible effects on things and people in case of error, but also with reference to their fundamental positive impact on the environment and on ethical issues.