人工智能倫理的挑戰與反思：文獻分析

Abstract

人工智能(Artificial Intelligence, AI)引領著大數據、物聯網以及工業 4.0 的整合應用，讓人們生活更加便利。然而也因AI演算法的優化、收斂與限制所導致在決策的偏見與缺失，加上數據取得之資訊?私和安全問題，以及可能發展成超越人類控制的超級智能、自主行動主體者的隱憂，衍生出種種的倫理議題成為AI發展是否需要共同準則與規範的疑慮。本文是搜尋以AI和倫理為關鍵詞的期刊論文後，針對2018至2019年的文獻進行綜合剖析，探討通用人工智能(AGI)、倫理的自主武器、AI醫療、AI商業模式等應用領域倫理議題，並針對AI系統規範、倫理嵌入AI系統、歧視、偏見和犯罪等倫理治理對策做整理與探討。本文期能借鏡國際上針對AI 倫理準則制訂範疇的學術論述，作為我國AI科技治理策略之參考。

Artificial intelligence (AI) is leading the integrations of big data, Internet of Things, and the Industry 4.0, resulting in the rapid development of unprecedented AI in all domains. Meanwhile, various ethical issues in AI are dramatically increasing in different domains. For instance, biases and deficiencies caused by the optimization, convergence, and limitation of algorithms, coupled with autonomous agents, information privacy, and security issues, adding the potential for developing superintelligence beyond human control. This research aims to systematically review studies from 2018 to 2019 concerning ethical issues in AI. Through an exploratory content analysis, four research themes emerged, namely, artificial general intelligence (AGI), ethical lethal autonomous weapons system (LAWS), AI medical treatment, and AI business model. For each theme, moreover, critical comments were made to highlight its controversy in the literature. Furthermore, the content analysis yielded in- depth discussions on ethical governance topics, including (1) ethics against AI research and design, (2) ethics embedded in AI system, (3) detecting discrimination, prejudice and crime, and (4) applying Human-in-the-Loop (HITL). In sum, the comments made by this literature review may capture the scope of the global-wise AI ethical policies and guidance, which can be implemented by government as strategic tech-references.