摘要
Sora作为革命性的“类世界模型”,以镜像进化的逻辑发展,引领“天人智一”认知融合新纪元。在文生视频技术驱动下,文章以三组核心矛盾作为切入点,探讨Sora对未来传媒生态认知的潜在影响。研究发现,Sora引领我们从“真实世界美学”探索到“异世界美学”等,这标志着从标准化资源向个性化创新的转变。同时,其预示着传媒主体将从传统的半人工编辑转向语义智能体,传媒场景将更深层次地打破物理时空区隔,传媒产业价值将朝向AI自动化等方向发展。此外,文章通过对Sora官网发布的48个视频进行案例分析发现,“AI幻觉”问题开始显现。未来传媒从业者在角色方面,需从生产者转变为核查者;在应用方面,需重视编码与解码过程;在治理层面,需从人工治理逻辑转向机器治理逻辑。具体而言,“AI幻觉”的治理包括基于博弈理论的幻觉识别技术,利用多角色、多智能体,进行AI系统间的博弈与碰撞,有效识别和修正幻觉内容。
Sora,as a revolutionary"world-like model",evolves with the logic of mirror evolution,leading the new era of cognitive integration,which means heaven,human,and intelligence are united as"One".Driven by the revolution of text-to-video technology,this study targets three sets of core contradictions as entry points,delves into the potential impacts of Sora on the cognition of future media ecology.It finds that Sora leads the exploration from"otherworldly aesthetics"to"anti-physical aesthetics",signifying a shift from standardized resources to personalized innovation.Concurrently,it forecasts that the media subject will transit from traditional semi-manual editing to semantic intelligent agents,deeply breaking the physical spacetime barriers in media scenarios,and steering the media industry's value towards the direction of AI hardware.Furthermore,a case analysis of 48 videos released on the Sora official website unveiled the emergence of the AI hallucination issue.In the future,media practitioners are required to shift their roles from producers to verifiers.In applications,emphasis should be laid on the encoding and decoding processes;in governance,a transition from manual governance logic to machine governance logic is necessary.Specifically,illusion recognition technologies based on game theory,utilizing multiple roles and agents,are needed for the games and collisions among AI systems to effectively identify and correct hallucinatory content.
作者
张诗瑶
沈阳
ZHANG Shi-yao;SHEN Yang(School of Journalism and Communication,Tsinghua University,Beijing 100084,China)
出处
《编辑之友》
北大核心
2024年第6期53-60,共8页
Editorial Friend
基金
国家自然科学基金青年项目“面向人工智能生成内容的风险识别与治理策略研究”(72304290)
清华大学—北京清尚建筑装饰工程有限公司智慧场景创新设计联合研究院横向课题“医院建筑装配式知识图谱构建与大语言模型综合研究”(20222910003)。