Global Efforts Towards Establishing Safety Directives for Intelligent Systems; Review

Abstract

Intelligent systems have found their way into our lives in almost every conceivable field.There is hardly any area in which the possibilities and applications of machine learning are not thought of.In safety-critical systems, there is an urgent need to examine how a model generates a prediction and whether that response can be trusted.Furthermore, ethics must be considered in the practical development of AI systems to ensure a safe and secure application.Despite these requirements, due to the nature of deep learning, we are confronted with a black box.This disparity needs to be addressed using interpretability and explainable approaches to minimize potential bias and at the same time increase transparency, fairness, justice and inclusion.In order to enhance trust in intelligent systems -accountability, responsibility and robustness must be ensured as well.Appropriate policies and standards need to be put in place to enforce this in practice.We are facing a global challenge here; standards must be set not only at national but also at international level, and a common understanding of how to deal with AI on ethical and legal levels must be found.We provide an overview of efforts that are being made at national and international level by governments and global organizations.We discuss current and upcoming challenges and risks posed by intelligent systems considering ethical guidelines and legal frameworks.In particular, we examine and compare the classification of risk levels and mitigation strategies.To conclude we show the latest state of technical feasibility and possible certification to ensure safe, transparent and robust AI systems and give an outlook on possible certification approaches for safe AI systems meeting the proposed governance frameworks.

Ähnliche Arbeiten