生成式AI的融貫性法律治理——以生成式預訓練模型(GPT)為例

The Coherent Legal Governance of Generative AI: Taking the Generative Pre-Training Model (GPT) as an example

隨著大規模數據和算式的增長，以及算法的不斷優化。圍繞生成式AI的前景，存在著支持、反對、中立等不同態度，這些態度背後隱含著認知根源、經濟考量和權利思考。立足於法律3. 0的融貫性治理注重國家法律、行政規制和技術方案之間的融貫，為治理生成式AI提供了思路和方向。融貫性治理中的“融貫”，既是規範性融貫，強調法律規範的內在一致性和統一性，也是整體性融貫，強調將技術方案融入規範，進而使得技術方案與不同層級規範和引領這些規範的原則與價值得以融貫。在面對以GPT為代表的生成式AI時，可以嘗試將AI和區塊鏈作為技術方案對其進行治理，也可以通過自我規制和外在約束培育建設“有道德的”AI，還可以通過“市場+規則”助力生成式AI的發展。生成式AI所涉及的法律問題在現有的法律體系框架內基本能得到有效應對，對於它帶來的現實、急迫且法律沒有明確規定的問題，可以進行融貫性治理。

With the explosive growth of new technologies and resources, such as large-scale data sets, computing power and algorithms, the generative AI represented by GPT is developing rapidly. Around the prospect of generative AI, there are different attitudes such as support, opposition and neutrality, which imply cognitive roots, economic considerations and rights thinking. The coherent governance based on law 3. 0 focuses on the coherence of national laws, administrative regulations and technical schemes, which provides ideas and directions for how to govern generative AI. The“coherence”in coherent governance is not only normative coherence, emphasizing the inherent consistency and unity of legal norms, but also holistic coherence, emphasizing the integration of technical schemes into norms, so that the technical schemes can be integrated with different levels of norms and the principles and values that lead to these norms. In the face of generative AI represented by GPT, we can try to govern it through technical schemes, such as AI and block-chain , and build“ethical”AI through self-regulation and external constraints, and promote the development of generative AI through“market + rules”. The legal issues involved in generative AI can be effectively dealt with within the framework of the existing legal system framework, and the problems brought about by generative AI, which are realistic, urgent and not clearly defined by law, can be governed in a coherence manner.