人工智能发展的系统性风险治理:实践反思、理性建构与策略优化

Systemic Risk Governance for Artificial Intelligence Development:Practical Reflection,Rational Construction and Strategy Optimization

人工智能的发展正不断深化风险复制与传递的连锁反应,对国家治理体系和治理能力现代化产生巨大挑战。以类型化风险为规制对象、以技术性视角为治理场域、以价值论思路为防范策略的模式,可为人工智能发展风险的概念化以及治理提供养分。但是,由于缺乏理性层面的“元风险”建构,使得风险治理的策略研究之间无法在治理对象上实现价值通约,无法兼顾人工智能发展中的公共安全风险与国家安全风险,无法对内外部风险的复杂形态形成有效的治理合力。基于总体国家安全观的系统性思维,理性建构之于风险本体的人工智能发展系统性风险之概念,并从要素层、观察层和动态层三个方面探究系统性风险的内生机理,以此构成解释和应对人工智能发展风险问题的全新视角。为此,总体国家安全观视域下人工智能发展系统性风险的治理优化,应当以人工智能与系统间的高度关联性为考量,将“类型化风险”转换为“节点风险”以实现治理对象的价值通约,用“关联性”修正“功能性”以明确技术治理的体系化逻辑,用“系统衡平”融合“价值导向”以强化风险治理体系的系统韧性。

The development of artificial intelligence(AI)is causing increasing chain reactions of risk replication and transmission,constituting a great challenge to modernizing the national system and capacity for governance.The model featuring with different types of risks as the object of regulation,technical perspective as the field of governance,and axiological thinking as the preventive strategy provides perspectives for the conceptualization and governance of the risk of AI development.However,due to the lack of construct in"meta-risk"at the rational level,the research on different risk governance strategies finds it difficult to achieve axiological commensuration among different governance objects,accommo-date both public security risk and national security risk in the Al development,and form an effective governance synergy for the complex internal and external risks.Based on the system thinking of a holistic approach to national security,we should rationally construct the systemic risk concept for AI develop-ment based on the risk itself and explore the endogenous mechanism of systemic risk from the elemental,observational and dynamic perspectives so as to create a brand-new perspective for explaining and re-sponding to the risks of AI development.Therefore,the governance optimization for the systemic risk of Al development from the perspective of a holistic approach to national security should take into account the high correlation between AI and the system.As a result,"different types of risks"should be trans-ferred into"nodal risks"to achieve the axiological commensuration for the governance objects.The"correlation"should replace"functionality"to clarify the systematic logic of technology governance.And"system equity"should be integrated into the"value orientation"to strengthen the system resilience of therisk governance system.