首页|人工智能治理的算力维度:论算力互联互通

人工智能治理的算力维度:论算力互联互通

扫码查看
人工智能产业因ChatGPT、Sora等产品的出现而备受关注.为了促进和保障人工智能产业发展和技术创新,各国的人工智能立法活动也相继开启.在学界试图明确训练数据合理使用范围、算法安全风险预防等争议问题时,算力供给相关的法律问题未能得到充分重视.特别是有关我国算力互联互通的研究大多聚焦于技术标准层面,忽视了算力互联互通作为一项法律制度的可能性与必要性.结合算力的业务流程来看,在算力生产整合阶段,"算力"是指具有资源型财产特征的算力资源,由国家统一管理规划;在算力供应交易阶段,"算力"是指具有私人财产属性的算力信息服务.因此,算力互联互通制度并非强制性的算力共享制度,而是基于政府必要干预的算力骨干网的强制有效连接和基于市场导向的算力市场交易的制度体系.在算力基础设施有序建设、算力资源可访问以及算力市场化交易三项基本原则导向下,算力互联互通制度化的根本目标是实现算力资源从生产到使用各环节的自由流动.
Dimension of Computing Power in the Governance of Artificial Intelligence:On the Interconnection of Computing Power
The artificial intelligence industry has attracted attention due to products like ChatGPT and Sora.Legislation on artificial intelligence in various countries has been initiated to promote and safeguard industry development and technological innovation.While academic circles attempt to clarify controversial issues such as reasonable use of training data and prevention of algorithmic security risks,legal issues related to computing power supply have not received sufficient attention.Particularly,most studies on China's inter-connection of computing power focus on technical standards,neglecting the possibility and necessity of interconnection of computing power as a legal system.Considering the business processes of computing power,during the integration stage of computing power pro-duction,"computing power" refers to resource—based property characteristics managed and planned by the state;during the trading stage of computing power supply,"computing power" refers to computing power information services with private property attributes.Therefore,the interconnection of computing power is not a mandatory sharing system of computing power but a system based on the nec-essary intervention of the government,effectively connecting the backbone network of computing power and establishing a system of mar-ket-oriented computing power transactions.Guided by three basic principles—orderly construction of computing power infrastructure,accessibility of computing power resources,and market-oriented trading of computing power—the fundamental goal of institutionalizing computing power interconnection is to achieve free flow of computing power resources from production to use.

赵精武、周瑞珏

展开 >

北京航空航天大学 法学院,北京 100191

北京航空航天大学 计算机学院,北京 100191

算力互联互通 算力资源 算力信息服务 算力基础设施

国家社会科学基金重大项目

16ZDA075

2024

郑州大学学报(哲学社会科学版)
郑州大学

郑州大学学报(哲学社会科学版)

CSSCICHSSCD北大核心
影响因子:0.506
ISSN:1001-8204
年,卷(期):2024.57(4)