인공지능 윤리의 주요 쟁점과 필수 고려사항

Abstract

Recently numerous ethical principles and guidelines for Artificial intelligence (henceforth, AI) have been developed though there are still many contentions and criticisms on the validity and practicality of AI ethics. To address this issue, this paper proposed five kernel research questions on the ethical issues of AI, based on the author’s seminal hypothesis, ‘can AI be an ethical subject?’. They are as follows: i) What are the problems and discrepancies in the development phases of AI ethics, ii) How much is AI ethics valid and on what theoretical basis does it stand on?, iii) Are the ethical guidelines practical in the real world?, iv) Who is the target of AI ethics development and how is the content structure of AI Ethics like? Finally, v) what are the practical considerations of AI ethics education and development? The major conclusive arguments of this paper are: first, current ethical guidelines are superficial, ambiguous, and biased in many aspects so that they can hardly represent the diverse socio-cultural conditions and backgrounds. Second, utilitarianism, rather than virtue or deontological ethics, would be most suitable philosophical basis since it can balance and reflect the characteristics of the developer, the algorithm, and the user. Third, to close the gap between the theory and practice AI should be an ethical subject, rather than an explicit proxy of ethics. This paper argued that developing ‘ethical AI’ should be the eventual goal of AI ethics in responding the sheer changes of a future society. This paper’s discussion may open the room for arguments in the area of AI studies.