| 李春南,王山.生成式人工智能时代的语言安全：系统性风险与治理路径[J].国际安全研究,2024,(5):135-156 |
| 生成式人工智能时代的语言安全：系统性风险与治理路径 |
| Linguistic Security in the Generative AI Era: An Analysis of Systemic Risks and Governance Strategies |
| 修订日期：2024-07-24 |
| DOI：10.14093/j.cnki.cn10-1132/d.2024.05.006 |
| 中文关键词: 生成式人工智能 语言安全 风险 安全治理 |
| 英文关键词: generative AI, linguistic security, risk, security governance |
| 基金项目: |
|
| 摘要点击次数: 687 |
| 全文下载次数: 173 |
| 中文摘要: |
| 随着生成式人工智能(Generative AI)的快速发展和广泛应用，语言安全逐渐超越了信息和文化安全等传统国家安全界限，在语言技术、语言认知、语言行动等维度面临新的挑战。识别和分析生成式人工智能时代的语言安全，把握其理论内涵和风险生成机制，对预防和化解语言安全风险意义重大。本文探究生成式人工智能时代语言安全的理论内涵及其构成要素，基于“技术—认知—行为”的分析框架厘清该时代语言安全风险的生成机理，提出针对性的治理策略。生成式人工智能时代语言安全的理论内核和构成要素呈现出技术、主体与社会互嵌的特性。语言安全在生成式人工智能技术研发和应用的不同环节存在差异化和动态化的语言风险样态和逻辑。针对生成式人工智能带来的语言安全风险提出制定和实施高层次国家语言安全战略、建立生成式人工智能自主生态体系、提高该技术的透明度与可解释性、开发多语种模型、提升公众数字素养和语言安全意识等治理对策。 |
| 英文摘要: |
| The rapid advancement and widespread application of generative artificial intelligence (Generative AI) have redefined the scope of linguistic security, extending its significance beyond traditional concerns of information and cultural security. This paper critically examines the emerging challenges posed by Generative AI to linguistic security across the dimensions of language technology, cognition, and behavior. By delineating the theoretical underpinnings and risk-generating mechanisms of linguistic security in the era of Generative AI, this paper underscores the importance of a nuanced understanding for effective mitigation of these risks. Utilizing a “technology-cognition-action” framework, the paper elucidates the complex interplay of factors that constitute linguistic security risks in this context. The findings reveal that the core theoretical constructs and components of linguistic security in the Generative AI era are characterized by intricate interdependence among technological, agentic and societal dimensions. Furthermore, the paper constantly identifies the dynamic and differentiated nature of linguistic risks at various stages of Generative AI development and application. In response, the paper proposes a suite of governance strategies, including formulating and implementing a high-level national linguistic security strategy, creating an autonomous Generative AI ecosystem, enhancing transparency and interpretability in Generative AI technologies, developing multilingual models, and promoting digital literacy and linguistic security awareness among the public. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|