A systematic literature review on ai safety: Identifying trends, challenges and future directions

Abstract

This systematic literature review addresses the critical need for safe and trustworthy AI systems amidst the rapid advancements and ethical concerns surrounding artificial intelligence. It provides a structured overview of AI safety, emphasizing the significance of designing AI systems with a safety focus encompassing data management, model development, and deployment. The study underscores the necessity for AI systems to align with human values and operate within established ethical frameworks. The methodology involves a comprehensive review of existing research to identify trends, challenges, and future directions in AI safety. The review highlights the association of AI safety with model learning techniques, verification and validation methods, failure modes, and the management of AI autonomy. Key findings reveal major concerns in the field, including explainability, interpretability, robustness, reliability, fairness, bias, and adversarial attacks. The implications of this review point to the need for a complete safety framework that guides the development and implementation of AI systems, ensuring they do not inadvertently cause harm to humans. The findings suggest that AI regulation and governance should prioritize the development of mechanisms and standards that address the identified concerns, fostering the creation of AI systems that are not only advanced but also safe, reliable, and aligned with human values.