理解生成式AI:融通机器智能与人类智能的算法媒介
Understanding Generative AI:An Algorithmic Media for Melding Machine Intelligence and Human Intelligence
刘彧晗 1喻国明1
作者信息
- 1. 北京师范大学新闻传播学院,北京 100091
- 折叠
摘要
以媒介视角考察生成式AI是本文探讨的核心问题.研究尝试从媒介形态与特性、媒介化机制与路径方面为这一媒介技术提供较全面的价值阐释.研究认为,生成式AI具有独特的"微调"可供性,并赋予了"居"于机器智能与人类智能之"间"、连接机器智能要素与人类智能要素的媒介特性.生成式AI的媒介化力量,是借助机器的参数解构力和人类的标注牵引力实现的社会强媒介化过程,机器智能发挥解构社会秩序和执行生产过程的功能,人类智能则担任牵引社会方向和"赋魂"智力创新的角色.展望未来发展,生成式AI短期需解决数据契合度问题,完善模型;中期需解决使用契合度问题,培训提示工程;长期需解决社会契合度问题,适应政策市场.由此,生成式AI方可发挥最大的媒介化效应,迈向人机协同新文明.
Abstract
The examination of generative AI from a media perspective is the core focus of this paper,which attempts to provide a comprehensive value elucidation for this media technology from aspects of media form and characteristics,as well as the mechanisms and pathways of mediatization.This paper posits that generative AI represents the latest form of algorithmic media,endowed with a unique"fine-tuning"affordance;this affordance confers upon it the media characteristic of"dwelling"between machine intelligence and human intelligence,connecting elements of machine intelligence with those of human intelligence.The mediatization power of generative AI is realized through a process of strong social mediatization,leveraging the parameter deconstruction power of machines and the annotational traction power of humans.Machine intelligence plays a role in deconstructing social order and executing production processes,while human intelligence takes on the role of steering social direction and"ensouling"intellectual innovation.Looking towards the future,generative AI needs to address the issue of data fit in the short term by refining models;in the medium term,it needs to address the issue of usage fit by training prompt engineering;and in the long term,it needs to address the issue of social fit by adapting to policies and markets.Only by doing so can it maximize its mediatization effect and move towards a new civilization of human-machine synergy.
关键词
生成式AI/媒介特性/媒介化/复杂性范式/社会系统论Key words
generative AI/media characteristics/mediatization/complexity paradigm/social systems theory引用本文复制引用
出版年
2024
NETL
NSTL
万方数据