人工智能大模型发展带来的风险挑战和对策  

Risks,Challenges and Countermeasures Arising from the Development of Large Models of Artificial Intelligence

作　　者：徐峰[1] 赛秋玥 刘鑫怡 刘乾 Xu Feng

机构地区：[1]中国科学技术信息研究所人工智能发展研究中心 [2]中国科学技术信息研究所

出　　处：《学术前沿》2024年第13期72-78,共7页Frontiers

基　　金：科技创新2030——“新一代人工智能”重大项目研究成果之一,课题编号:2023ZD0121701。

摘　　要：近年来,以预训练大模型为代表的人工智能技术能力快速提升,也加速推动人工智能技术与科技、经济、社会发展深度融合,成为新一轮科技革命和产业变革的重要驱动力量。作为一项新兴技术,人工智能的“双刃剑”效应使得技术在快速进步和应用的同时,也引发了全球对其可能带来各类风险挑战的担忧。特别是随着大模型技术的突破,人工智能可能带来的风险挑战发生了一些新的变化。因此,需要在充分把握人工智能技术发展规律的基础上,从政策法规、技术能力、标准规范等多个方面采取针对性的治理之策,深化全球协作,共同应对风险挑战。In recent years,the capability of artificial intelligence technology represented by the pretraining large model is improved rapidly Artificial intelligence technology is deeply integrated with the development of science and technology,economy and society,which has become the important driving force of the new round of science and technology revolution and industrial transformation.At the same time,the"double-edged sword"effect of artificial intelligence as a new technology causes the global concern to the risks and challenges it may bring,with the technology rapidly advance and apply.Especially with the breakthrough of large model technology,the risks and challenges of artificial intelligence have brought some new changes.Therefore,we should take targeted governance countermeasures from policy,regulation,technical capability,standard and so on the basis of grasping the law of the development of artificial intelligence technology,and actively deal with them under the global cooperation.

关 键 词：人工智能 预训练大模型 风险挑战 治理对策 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]