卫生领域人工智能的伦理与治理:多模态大模型指南  

Ethics and governance of artificial intelligence for health:guidance on large multi-modal models

作　　者：王玥(译)[1] 宋雅鑫(译) 王艺霏(译) 于莲(审校) 王晶(审校)[3] WANG Yue;SONG Yaxin;WANG Yifei;YU Lian;WANG Jing(School of Law,Xi’an Jiaotong University,Xi’an 710049,China;School of Public Health,Xi’an Jiaotong University,Xi’an 710061,China;Beijing Hospital of Traditional Chinese Medicine,Capital Medical University,Beijing 100010,China)

机构地区：[1]西安交通大学法学院,陕西西安710049 [2]西安交通大学公共卫生学院,陕西西安710061 [3]首都医科大学附属北京中医医院,北京100010

出　　处：《中国医学伦理学》2024年第9期1001-1022,共22页Chinese Medical Ethics

摘　　要：2024年,世界卫生组织发布了“Ethics and governance of artificial intelligence for health.Guidance on large multi-modal models”,将其翻译成中文《卫生领域人工智能的伦理与治理:多模态大模型指南》供中国的同仁参阅,协助规划与卫生领域多模态大模型有关的益处和挑战,并为适当开发、提供和使用多模态大模型提供政策和实践方面的指导。世界卫生组织咨询了20位人工智能领域的顶尖专家,他们确定了在卫生领域使用人工智能的潜在益处和潜在风险,并发布了以协商方式达成一致的六项原则,供正在使用人工智能的政府、开发者和提供者在制定政策和实践时考虑。指南提供了与指导原则相一致的企业内部、政府和国际合作的治理建议,指南的基础是考虑到人类使用卫生领域生成式人工智能独特方式的指导原则和治理建议。生成式人工智能是算法在可用于生成新内容的数据集上进行训练的一种人工智能技术。指南针对其中一种类型的生成式人工智能,即多模态大模型,这种模型可以接受一种或多种类型的数据输入,并产生不局限于输入算法的数据类型的多种输出。据预测,多模态大模型将广泛应用于医疗保健、科学研究、公共卫生和药物开发等领域。多模态大模型也被称为“通用基础模型”,尽管尚未证实多模态大模型能否完成各种任务和目的。In 2024,the World Health Organization(WHO)released“Ethics and governance of artificial intelligence for health.Guidance on large multi-modal models”,which was translated into Chinese for Chinese readers to read and think about,assist in planning the benefits and challenges related to large multimodal models for health,as well as provide guidance on policy and practice for the appropriate development,provision,and use of large multi-modal models.The WHO consulted 20 leading experts in the field of AI,who identified potential benefits and risks of using AI for health,as well as published six principles agreed upon in a consultative manner for governments,developers,and providers using AI to consider when formulating policies and practices.The guidelines provide governance recommendations that are consistent with the guiding principles for internal enterprise,governments,and international cooperations,which are based on the guiding principles and governance recommendations that consider the unique ways in which humans use AI generated content(AIGC)for health.AIGC is an AI technology in which algorithms are trained on data sets that can be used to generate new content.The guidelines target one type of AIGC,namely the large multi-modal models abbreviated as“LMM”,which can accept one or more types of data inputs and generate multiple outputs of data types that are not limited to the input algorithm.It is predicted that LMM will be widely used in various fields,such as healthcare,scientific research,public health,and drug development.LMM are also called“universal base models”,although it has not yet been proven that LMM can complete a variety of tasks and purposes.

关 键 词：卫生领域人工智能 多模态大模型 通用基础模型 伦理与治理 

分 类 号：R-052[医药卫生]