“世界模拟”的拟像迷思——基于通用视觉大模型技术的哲学反思  

The Simulacra Myth of“World Simulation”—Philosophical Reflection Based on General Large Vision Models

作　　者：吴静 WU Jing

机构地区：[1]南京师范大学公共管理学院 [2]南京师范大学数字与人文研究中心

出　　处：《南通大学学报（社会科学版）》2024年第3期20-30,159,共12页Journal of Nantong University:Social Sciences Edition

摘　　要：随着通用视觉大模型技术的迅速发展,对人工智能技术底层逻辑的哲学反思变得刻不容缓。生成式人工智能文生视频、文生图像现象的背后,是数字技术借由算法公理化逻辑所营造出的普世视觉景观,这种视觉景观消解了真实与虚拟之间的边界,在本质上与一种通过数字技术而布展的知识生产权力具有同构性。基于数据预训练和投喂的通用视觉大模型,其知识生产中存在着数据“通用”性与模型“泛化”的张力,大模型泛化能力的提高意味着其所依赖的数据来源愈加具有普遍性和公理性,由此在技术无意识层面形成一种代表数字普遍理性的公共知识体系。为此应重新思考虚拟与现实之间的边界问题,在技术设计关注差异要素的基础上,探索人工智能时代人机交互的可能前景。General large vision models growing rapidlyr philosophical reflections on the underlying logic of artificial inteligence technology should be made with no delays.Behind the phenomenon of‘texts to videos'and‘texts to images'of generative artificial intelligence is the universal spectacle of vision created by digital technology through the axiomatic logic of algorithms,which dissolves the boundary between reality and virtuality,homogeneous in essence with the power of knowledge production,As general large vision models based on pre-trained and fed data,there is a tension between the"universality"of the data and the"generalization"of the model during the process of knowledge production.The improvement of the generalization ability of the large model means that the data sources it relies on are increasingly universal and axiomatic,thus forming a public knowledge system representing digital universal rationality at the technical unconscious level.It is therefore necessary to rethink the boundary between reality and virtuality for the exploration into possible prospects of human-technique interaction in the era of artificial intelligence on the basis of focusing on diferent elements in technical design,

关 键 词：人工智能 文生视频 大模型 泛化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] N02[自动化与计算机技术—控制科学与工程]