关键词: 生成式人工智能, 风险语义图像, 技术风险

Abstract: While generative AI brings technological breakthroughs and improves efficiency,it has also evolved into an invisible technological risk in the public view.This research is grounded in the perspective of the semantic imagery of technological risks and employs computational text analysis,integrating LDA topic modeling,word vector models,and semantic networks to analyze discussions on AI risks within online communities(Zhihu).Results show that public perceptions of AI risks focus on the short-term risks of ChatGPT in information dissemination and the long-term risks of AI in human society.Public risk perception exhibits an individualization trend,with AI being viewed as a low-harm risk of voluntary involvement,while cognitively triggering public uncertainty.The identification of risk types by individuals is focused on national security risks and personal economic risks,presenting a cognitive framework that combines macro-politics with micro-individual interests.The public mainly adopts two coping strategies:a knowledge-centered self-learning strategy and an industry-centered collective response strategy,ultimately forming four risk semantic images including technological regulation,technological governance,survival adaptation,and systemic resonance.

Key words: generative AI, risk semantic imagery, technological risk