乌合之众的超级节点?AI大模型使用的人机网络结构分析  

Supernodes of the Crowd? Human-Machine network structure used by AI large models

作　　者：张洪忠 王彦博 任吴炯 刘绍强 Zhang Hongzhong;Wang Yanbo;Ren Wujiong;Liu Shaoqiang(School of Journalism and Communication,Beijing Normal University;School of Educational Technology,Northwest Normal University)

机构地区：[1]北京师范大学新闻传播学院,北京100875 [2]西北师范大学教育技术学院

出　　处：《新闻界》2023年第10期12-19,共8页Journalism and Mass Communication

基　　金：国家社会科学基金重点项目“人工智能技术背景下加快国际传播能力建设研究”(22AZD072);西北师范大学青年教师科研能力提升计划项目(NWNUSKQN2021-30)。

摘　　要：大模型超越“图灵测试”标准,使得机器开始具备接近“人”的行为特征,深刻影响人机混合的网络结构。本文分析大模型在信息传播网络中扮演中心节点角色:大模型及基于大模型的应用直接与大量用户建立“一对一”连接关系,进而推动网络结构向“再中心化”演变,成为一个超级节点。大模型通过学习网络“众人”知识提供新的经验数据,其本质是个体与算法支配下的“众人”互动。随着大模型深入传播实践,本文思考这种趋势是否会带来传受双方的乌合之众,并进一步从传受双方角度讨论了大模型网络结构的可能影响。The use of large models surpasses the“Turing Test”standard,enabling machines to exhibit behavior characteristics close to those of humans,profoundly influencing the network structure of human-machine hybrid interactions.This article analyzes how large models play a supernodesrole in information dissemination networks.Large models and applications built upon them establish direct“one-to-one”connections with a vast number of users,thereby driving the network structure towards“recentralization”.Thereby,large models serving as supernodes.Large models provide new experiential data by learning from the collective knowledge of the“crowd”,essentially representing interactions between individuals and algorithms.As large models continue to spread in practical applications,this article contemplates whether this might lead to a convergence of the crowd on both the transmitting and receiving ends.Furthermore,this article discusses the potential impact of large model network structures from the perspectives of both transmitters and receivers.

关 键 词：大模型 机器行为 超级节点 网络结构 乌合之众 

分 类 号：G206[文化科学—传播学]