跨模态AI、人机协作与创造力:以文本生成图像模型为中心的技术重思  

Cross-Modal AI,Human-Machine Collaboration and Creativity:Rethinking the Technology with Text-to-Image Generation Modals as the Core

作　　者：杨俊蕾[1] 郑丹路 Yang Junlei;Zheng Danlu

机构地区：[1]复旦大学中国语言文学系,上海200433

出　　处：《中国社会科学评价》2024年第2期45-53,158,共10页China Social Science Review

基　　金：教育部规划基金项目“数字人文赋能中国电影理论创新研究”(23YJA760105)阶段性成果。

摘　　要：AIGC领域中的文本生成图像模型主要基于两个技术环节:文图关联和图像生成。当前的数智化艺术实践以文本生成图像模型为工具,通过基准偏移、提供巨量数据点位的方式对艺术创作进行训练和模拟,模型性能和用户体验双双呈现出高速发展的态势。包括文生图模型在内的AIGC应用程序将人机协作这一薪新的创作范式带入艺术创作领域,为了获得机器的协助,人类须调整自我,向机器思维靠拢,同时人在这个过程中也达成了对自身思维方式的重思。现有的AI生成技术或许尚未使文生图模型获得自主创造力,未来的艺术创作或许不再呈现为当下人与机器的竞争较量,而是呈现为“人十机器”与“人十机器”之间的竞争较量。人与机器作为一个新的整体,成为创造力的源泉。In the realm of AI-generated content creation(AIGC),text-to-image generation models primarily depend on two technological processes:text-image correlation and image generation.AIGC applications,including text-to-image models,introduce a novel paradigm of humanmachine collaboration into the domain of artistic creation.To harness the assistance of machines,humans must recalibrate their approach,aligning closer to machine-like thinking,while concurrently reevaluating therir own cognitive processesA.lthough current Al generation technologies may not yet endow text-to-image models with autonomous creativity,the future of artistic creation may no longer be characterized by a competition between humans and machines,but rather a contest between permutations of"human+machine"collaborations.Together,humans and machines form a new entity,becoming a wellspring of creativity.

关 键 词：跨模态AI 人机协作 创造力 文本生成图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]