人工智能生成内容的潜在风险及治理对策

doi:10.11959/j.issn.2096-6652.202332

智能科学与技术学报 ›› 2023, Vol. 5 ›› Issue (3): 415-423.doi: 10.11959/j.issn.2096-6652.202332

• 专题：扩散模型和人工智能内容生成 • 上一篇下一篇

人工智能生成内容的潜在风险及治理对策

李亚玲¹^,², 覃缘琪¹^,², 魏阙³

¹ 之江实验室智能社会治理研究中心，浙江杭州 311121
² 浙江省哲学社会科学试点实验室——之江实验室智能社会治理实验室，浙江杭州 311121
³ 之江实验室发展研究中心，浙江杭州 311121

修回日期:2023-07-25 出版日期:2023-09-01 发布日期:2023-09-26
作者简介:李亚玲（1991- ），女，博士，之江实验室智能社会治理研究中心助理研究员，主要研究方向为技术治理和数字经济
覃缘琪（1992- ），女，之江实验室智能社会治理研究中心助理研究员，主要研究方向为智能交通控制和智能社会治理
魏阙（1986- ），男，博士，之江实验室发展研究中心助理研究员，主要研究方向为科技创新生态
基金资助:
国家重点研发计划(2022YFC3303103);浙江省科技厅软科学计划研究项目(2023C25074);中共浙江省委政法委员会浙江省法学会2023年度法学研究课题重点项目(2023NA02);浙江省科学技术情报学会科研项目(2023qbxh006);杭州市余杭区社会科学研究常规资助课题(Yhsk23C01)

Potential risks and governance strategies of artificial intelligence generated content technology

Yaling LI¹^,², Yuanqi QIN¹^,², Que WEI³

¹ Research Center for Intelligent Society and Governance, Zhejiang Laboratory, Hangzhou 311121, China
² Zhejiang Labs for Philosophy and Social Sciences: Laboratory for Intelligent Society and Governance in Zhejiang Laboratory, Hangzhou 311121, China
³ Center for Science and Technology Strategy, Zhejiang Laboratory, Hangzhou 311121, China

Revised:2023-07-25 Online:2023-09-01 Published:2023-09-26
Supported by:
The National Key Research and Development Program of China(2022YFC3303103);Zhejiang Provincial Department of Science and Technology Soft Science Research Program(2023C25074);Research Project of Zhejiang Law Society(2023NA02);Research Project of Zhejiang Society for Scientific and Technical Information(2023qbxh006);Social Science Research Project of Yuhang District(Yhsk23C01)

摘要/Abstract

摘要：

人工智能生成内容技术正重塑数字内容的生产方式和消费模式。基于人工智能生成内容技术的基本概念和技术发展历程，重点研究了人工智能生成内容引发的数据泄露、算法霸权、数字鸿沟、知识产权保护、就业冲击等潜在风险和治理挑战。在梳理了国外发达国家的典型治理路径和治理经验后，概括提出了我国面对人工智能生成内容技术引发的潜在风险的应对措施和发展建议。具体包括坚持安全与发展并重，做好统筹布局；完善人工智能技术治理规则体系；加强人工智能技术研发全要素投入；加快人工智能治理技术研发；构建全方位多层次的技术治理体系。

关键词: 人工智能生成内容, 风险, 技术治理, ChatGPT

Abstract:

Artificial intelligence generated content technology is reshaping the production and consumption patterns of digital content.The research was based on the basic concepts and technological development process of artificial intelligence generated content technology, focusing on exploring ethical risks and governance challenges such as privacy leakage, algorithm hegemony, digital divide, intellectual property protection and employment impact caused by artificial intelligence generated content.After reviewing the typical governance paths and experiences of developed countries abroad, this paper summarized and proposed measures and development suggestions for China to address the potential risks of AI generated content.One was to prioritize development and coordinate layout; Second was to strengthen the full factor investment in the research and development of artificial intelligence technology; Third was to improve the governance rule system of artificial intelligence technology; Fourth was to accelerate the research and development of artificial intelligence governance technology.

Key words: artificial intelligence generated content, risk, technology governance, ChatGPT

中图分类号:

T-01

李亚玲,覃缘琪,魏阙. 人工智能生成内容的潜在风险及治理对策[J]. 智能科学与技术学报, 2023, 5(3): 415-423.

Yaling LI,Yuanqi QIN,Que WEI. Potential risks and governance strategies of artificial intelligence generated content technology[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(3): 415-423.

图/表 1

参考文献 30

[1]	郁建兴, 刘宇轩, 吴超 . 人工智能大模型的变革与治理[J]. 中国行政管理, 2023,39(4): 6-13.
	YU J X , LIU Y X , WU C . The social transformation and governance of large generative artificial intelligence models[J]. Chinese Public Administration, 2023,39(4): 6-13.
[2]	张凌寒, 于琳 . 从传统治理到敏捷治理:生成式人工智能的治理范式革新[J]. 电子政务,已录用, 2023.
	ZHANG L H , YU L . From traditional governance to agile governance:innovation in the governance paradigm of generative artificial intelligence[J]. E-Government,accepted, 2023.
[3]	宋恺, 屈蕾蕾, 杨萌科 . 生成式人工智能的治理策略研究[J]. 信息通信技术与政策, 2023,49(7): 83-88.
	SONG K , QU L L , YANG M K . Research on governance strategy of generative artificial intelligence[J]. Information and Communications Technology and Policy, 2023,49(7): 83-88.
[4]	于水, 范德志 . 新一代人工智能(ChatGPT)的主要特征、社会风险及其治理路径[J]. 大连理工大学学报(社会科学版),已录用, 2023.
	YU S , FAN D Z . The main characteristics,social risks and governance paths of the new generation of artificial intelligence(ChatGPT)[J]. Journal of Dalian University of Technology(Social Sciences),accepted, 2023.
[5]	张欣 . 生成式人工智能的数据风险与治理路径[J]. 法律科学(西北政法大学学报), 2023(5): 1-13.
	ZHANG X . The data risk and governance path of generative artificial intelligence[J]. Science of Law(Journal of Northwest University of Political Science and Law), 2023(5): 1-13.
[6]	WU J , GAN W , CHEN Z ,et al. AI-generated content (AIGC):a survey[EB]. arXiv preprint, 2023,arXiv:2304.06632.
[7]	YOUNG S , BLOOTHOOFT G . Corpus-based methods in language and speech processing[J]. Springer Science ＆ Business Media, 1997.
[8]	REYNOLDS D , LI S Z , JAIN A . Gaussian mixture models[M]// Encyclopedia of Biometrics. Boston.MA: Springer, 2009: 659-663.
[9]	BENGIO Y , DUCHARME R , VINCENT P . A neural probabilistic language model[J]. Advances in Neural Information Processing Systems, 2000.
[10]	MIKOLOV T , KARAFIAT M , BURGET L ,et al. Recurrent neural network based language model[J]. In Interspeech, 2010,2(3): 1045-1048.
[11]	GRAVES A . Long short-term memory[M]// Supervised Sequence Labelling with Recurrent Neural Networks. Berlin,Heidelberg: Springer, 2012: 37-45.
[12]	DEY R , SALEM F M . Gate-variants of gated recurrent unit (GRU) neural networks[C]// Proceedings of 2017 IEEE 60th International Midwest Symposium on Circuits and Systems (MWSCAS). Piscataway:IEEE Press, 2017: 1597-1600.
[13]	EFROS A A , LEUNG T K . Texture synthesis by non-parametric sampling[C]// Proceedings of the Seventh IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2002: 1033-1038.
[14]	HECKBERT P S . Survey of texture mapping[J]. IEEE Computer Graphics and Applications, 1986,6(11): 56-67.
[15]	CRESWELL A , WHITE T , DUMOULIN V ,et al. Generative adversarial networks:an overview[J]. IEEE Signal Processing Magazine, 2018,35(1): 53-65.
[16]	SONG Y , ERMON S . Generative modeling by estimating gradients of the data distribution[J]. Advances in Neural Information Processing Systems, 2019,32.
[17]	VASWANI A , SHAZEER N , PARMAR N ,et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017,30.
[18]	DOSOVITSKIY A , BEYER L , KOLESNIKOV A ,et al. An image is worth 16x16 words:transformers for image recognition at scale[EB]. arXiv preprint, 2020,arXiv:2010.11929.
[19]	LIU Z , LIN Y T , CAO Y ,et al. Swin transformer:hierarchical vision transformer using shifted windows[C]// Proceedings of 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway:IEEE Press, 2022: 9992-10002.
[20]	RADFORD A , KIM J W , HALLACY C ,et al. Learning transferable visual models from natural language supervision[C]// Proceedings of International Conference on Machine Learning.New York:PMLR. 2021: 8748-8763.
[21]	CAO Y , LI S , LIU Y ,et al. A comprehensive survey of AI-generated content (AIGC):a history of generative AI from GAN to ChatGPT[EB]. arXiv preprint, 2023,arXiv:2303.04226.
[22]	OUYANG L , WU J , JIANG X ,et al. Training language models to follow instructions with human feedback[EB]. arXiv preprint, 2022:arXiv:2203.02155.
[23]	TAN F X , YAN P F , GUAN X P . Deep reinforcement learning:from Qlearning to deep Q-learning[M]// Neural Information Processing. Cham: Springer International Publishing, 2017: 475-483.
[24]	ROMBACH R , BLATTMANN A , LORENZ D ,et al. High-resolution image synthesis with latent diffusion models[C]// Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2022: 10674-10685.
[25]	CARLINI N , IPPOLITO D , JAGIELSKI M ,et al. Quantifying memorization across neural language models[EB]. arXiv preprint, 2022,arXiv:2202.07646.
[26]	CARLINI N , HAYES J , NASR M ,et al. Extracting training data from diffusion models[EB]. arXiv preprint, 2023,arXiv:2301.13188.
[27]	BORJI A . A categorical archive of ChatGPT failures[EB]. arXiv preprint, 2023,arXiv:2302.03494.
[28]	SCHUHMANN C , VENCU R , BEAUMONT R ,et al. Laion-400m:open dataset of clip-filtered 400 million image-text pairs[EB]. arXiv preprint, 2021,arXiv:2111.02114.
[29]	周慧珺, 邹文博 . 数字化转型背景下数字鸿沟的现状、影响与应对策略[J]. 当代经济管理, 2023,45(3): 60-67.
	ZHOU H J , ZOU W B . The present situation,influence and countermeasures of the digital divide under the background of digital transformation[J]. Contemporary Economic Management, 2023,45(3): 60-67.
[30]	李春芳, 黄涛 . 受众视角下人工智能生成物的版权保护探析[J]. 华南理工大学学报(社会科学版), 2022,24(1): 24-32.
	LI C F , HUANG T . Copyright protection of artificial intelligence products from the perspective of audience[J]. Journal of South China University of Technology(Social Science Edition), 2022,24(1): 24-32.

模型类别	模型名称	发布机构	发布时间	应用领域
	LaMDA	Google	2021年5月	文本生成、对话
	ChatGPT	Open AI	2022年11月	文本生成、对话
	LLaMA	Meta	2023年2月	文本生成、对话
单模态	Alpaca	斯坦福大学	2023年3月	文本生成、对话
	文心一言	百度	2023年3月	文本生成、对话
	星火	科大讯飞	2023年5月	文本生成、对话
	StyleGAN	NVIDIA	2018年12月	图像生成
	DVD-GAN	DeepMind	2019年7月	视频生成
	DALL·E 2	Open AI	2022年7月	从文本生成图像
	Stable Diffusion	Stability.AI	2022年8月	从文本生成图像
多模态	GPT-4	Open AI	2023年3月	支持图片和文本的多模态输入
	DeepFloyd IF	Stability.AI	2023年4月	从文本生成图像
	ImageBind	Meta	2023年5月	打通视觉、红外图像、文本、音频、深度信息、动作捕捉6种模态

人工智能生成内容的潜在风险及治理对策

Potential risks and governance strategies of artificial intelligence generated content technology

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 1

参考文献 30

相关文章 0

Metrics

推荐阅读 0