by Marcello Condemi
Artificial intelligence - as a technology that makes it possible to simulate (through the creation and
the application of integrated algorithms in a dynamic computing environment) intelligence processes
human beings, enabling them to think and act like human beings - is capable of generating depending on
of the designed algorithm is an extraordinary capacity for calculation, assimilation and
interpretation of enormous masses of data completely unimaginable to human capacity, both the
violation of fundamental individual rights, going so far as to exploit its
habits, expectations and prejudices, to push him to make driven choices. Hence - beyond
of the already wide and useful application of AI in many areas of science and business (enterprise,
finance, medicine, communications, information, etc.) - the reasons for the concerns it
arouses when left to free enterprise and the consequent demand from
by jurists and scientists, to draw up rules to prevent their unwise use, detrimental to the rights
fundamental rights of the individual and as a cascade of harmonious social coexistence (e.g. the
possible manipulation of elections). Motivated by these concerns, the Council and the
European Parliament, first in the world, reached an agreement last December
provisional on the proposal on the introduction of harmonised rules on artificial intelligence
(id est: Artificial Intelligence Regulation), aimed at ensuring that AI systems used
in the EU are safe and respect fundamental rights and founding values. The agreement,
as a potential benchmark for other jurisdictions, stimulates the adoption of AI by preventing its
unacceptable applications; the agreement in particular, following a “risk-based” approach,
outlines regulatory cases (also backed by heavy fines)
proportionally more stringent as the riskiness of the AI model used increases assuming
(i) “limited risk” systems subject only to transparency obligations; (ii) “high risk” systems”
authorised where certain requirements are met and obligations are fulfilled; iii)
“unacceptable” and therefore prohibited, such as, for example, applications that aim to
behavioural manipulation, emotion recognition, forms of categorisation
biometrics, etc.. It is hoped that the process of finalising this body of legislation will be completed as soon as possible.
as quickly as possible, in order to prevent human intelligence, which is then the creator and
responsible for the AI application degenerations that are intended to be prevented, produce monsters
even worse than what mankind itself is sometimes capable of generating.
Published in the newspaper “Il Tempo” on 19 January 2024
