Ethics and Safety in AI Fine-Tuning

Bohdan Kovalevskyi ()

Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023, 2024, vol. 1, issue 1, 259-267

Abstract: This paper examines the ethical implications and technical challenges of AI model fine-tuning, focusing on the dichotomy between aligned and unaligned models. Through analysis of current practices and emerging frameworks, we explore how fine-tuning can simultaneously enhance model performance and introduce potential risks. The study investigates the mathematical foundations of fine-tuning processes, ethical considerations in model alignment, and the challenges of balancing innovation with safety. We propose a composable alignment approach that maintains core ethical principles while allowing context-sensitive applications. The paper also evaluates existing regulatory frameworks and their effectiveness in governing AI development, suggesting mechanisms for oversight. Our findings emphasize the need for adaptive alignment strategies and global collaboration in establishing ethical standards for AI alignment, while highlighting the importance of maintaining flexibility across different cultural and application contexts.

Keywords: artificial intelligence; model fine-tuning; AI alignment; ethical AI; regulatory frameworks; composable alignment; AI safety; model bias; AI governance (search for similar items in EconPapers)
Date: 2024
References: Add references at CitEc
Citations:

Downloads: (external link)
https://newjaigs.com/index.php/JAIGS/article/view/275 (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:das:njaigs:v:1:y:2024:i:1:p:259-267:id:275

Access Statistics for this article

Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023 is currently edited by Justyna Żywiołek

More articles in Journal of Artificial Intelligence General science (JAIGS) ISSN:3006-4023 from Open Knowledge
Bibliographic data for series maintained by Open Knowledge ().