人工智能時代的規範延遲：問題與應對——基於法哲學視角的考察

人工智能技術狂飆背後潛藏著多重結構性挑戰，它使人類現有的法律、倫理等規範跟進速度遲緩，難以穿透人工智能設計、生產和應用的全鏈條進行監管，由此產生調整滯後的規範治理時間差，甚或治理真空，此即所謂''規範延遲''現象。這需要我們審視有關''人工智能作為對象的治理''和''借助人工智能的治理''這兩種治理類型，考察它們在人類現實生活中展現的相互交錯關係，設計既尊重人類基本價值、又能發揮人工智能治理潛在效能的''人主機輔''人—機規範協同''的秩序與規範世界圖景。如果未來強人工智能（通用人工智能及超人工智能）在模擬人類的''意識''上真正實現了實質性的技術突破，則如何讓人工智能變得更加安全便應當成為人類社會在發展和使用人工智能技術時的首要考量因素。我們應建立一種''安全防禦''的新型治理範式，有必要將''人類主導權原則''和人類''不可讓渡領域''以代碼形式直接強制寫入所有強人工智能系統的底層架構，確保最終撤銷權與解釋權屬於人類本體。

Behind the advance of AI technology lie multiple structural challenges. They slow the pace at which existing human norms, such as law and ethics, can keep up, and make it difficult for regulation to penetrate the entire chain of AI design, production and application. This creates a temporal gap in normative governance that fails to adjust in time—or even a governance vacuum—the phenomenon of“normative lag”. This requires us to examine the two types of AI governance, namely“Governance of AI”and“Governance by AI”, to investigate the interwoven relationship they display in human social practice, and to design a vision of an ordered and normative world that is“human-led and machine-assisted”and based on“human-machine normative coordination”, one that both respects fundamental human values and brings into play the potential efficacy of AI governance. If, in the future, strong AI (artificial general intelligence, artificial superintelligence) achieves a substantive technological breakthrough in simulating human“consciousness”, then“how to make AI safer”should become the foremost consideration for human society in developing and using AI technology. A new governance paradigm of“safety defense”should be established, and it is necessary to embed the“principle of human primacy”and the human“inalienable domain”directly and compulsorily into the underlying architecture of all strong AI systems via coding, thereby ensuring that the ultimate right of revocation and the right of interpretation remain vested in the human entity itself.