“媒介机器学习”冲击下理解/重构媒介的技术发展思考——基于对奈飞匹配剪辑算法的考察

On Technological Development in Understanding/Reconstructing Media under the Impact of“Media ML”:An Analysis of Netflix Matching Clip Algorithm

流媒体内容平台奈飞于2022年底开始转型部署“奈飞机器学习平台”(MLP),以匹配剪辑算法为代表,基于机器学习和计算机视觉(人工智能技术)实现多模态内容理解。“奈飞机器学习平台”将应用于原创剧集内容制作和发行全流程,赋能面向订阅会员的个性化内容服务,预示着流媒体将从传统内容制作、发行平台转型成为“媒介机器学习”基础设施。伴随生成式人工智能引发的应用和监管问题,当前“奈飞机器学习平台”又为创意产业带来新一轮“媒介机器学习”冲击。不同于生成式人工智能日趋成熟的提示词研究和“拼装”美学特征,以奈飞匹配剪辑算法为代表的“媒介机器学习”在“规模理论”影响下对“数据洪流”进行特征提取,出于明确的营销意图向用户输出定制化视觉体验,以此强化推荐系统效能、提高用户留存率。回顾匹配剪辑在先锋艺术时期与计算机视觉领域的算法演进,当前奈飞匹配剪辑算法已然脱离构成主义趣味,在流媒体深耕“媒介机器学习”的转型策略下展露出理解/重构“媒介”的“算法文化”。在此趋势下,“媒介机器学习”虽然赋能流媒体平台快速迭代其创意理念,但同时也促使内容创作生态彻底融入庞大的算法推荐系统。

In the late 2022,Netflix began transitioning and deploying the Netflix Machine Learning Platform(MLP),with the matching clip algorithm as a representative example,leveraging machine learning and computer vision(artificial intelligence technology)to achieve multimodal content understanding.The Netflix MLP is set to be applied across the entire production and distribution process of original series content,empowering personalized content services for its subscribers.This shift signals the transformation of streaming platforms from traditional content production and distribution entities into Media ML infrastructures.Alongside the challenges caused by generative AI,the Netflix MLP introduces a new round of Media ML impact on the creative industries.Different from prompt-based research and assemblage aesthetic characteristics of generative AI,the Media ML represented by Netflix matching clip algorithm extracts features from the data deluge under the influence of scaling law and outputs customized visual experiences to users with clear marketing intentions,thereby enhancing recommendation system efficiency and improving user retention.Looking back at the evolution of matching clip from the period of avant-garde art to the field of computer vision,Netflix matching clip algorithm detoured the constructivist interests,revealing an Algorithmic Culture that seeks to understand and reframe media within the context of Netflix's strategies towards Netflix MLP.