The Negotiated Image

Title: The Negotiated Image

Year: 2025

Technique: AI-coded algorithmic generative moving image

Filming “AI’s Daily Generation” as a routine negotiation, the image is not the end point of expression but a byproduct of that negotiation: between prior data and current instructions, the system uses probabilistic experimentation, settling on the point of minimal error rather than the point of maximal meaning.

I collaborate with LLM through experimental programming to generate program architecture and shader algorithms. I also provide sampleable latent material through previously trained GANs and Diffusion models. The work itself is a real-time generative system, and the nine-and-a-half-minute video is merely a pixelated souvenir of a single run.

Generation is not about “what is expressed,” but rather “what is permitted and what is prohibited.” Authorship is merely a temporary consensus reached between two sets of constraints in a single frame. Pixels are merely a lossy translation of latent space; they are not “images of thought” but traces left by rules in time.

Exhibition:

• 2025 The Negotiated Image, Everyday Generation, Tianjin Academy of Fine Arts.
• 2025 The Negotiated Image, ARTE LAGUNA PRIZE Exhibition Shanghai.

作品名称：被协商的图像

创作年份：2025

作品类型：基于AI编码的算法生成影像

把“AI的日常生成”拍成一场例行的协商，图像并非表达的终点，而是协商的副产物：在先验数据与当下指令之间，系统以概率试探，在误差最小处而非意义最大处暂泊。

我以实验编程方式与大语言模型协作，生成程序架构和着色器算法。亦通过历史训练的生成对抗网络与扩散模型，提供可采样的潜在物质。作品本体是实时生成系统，九分半视频不过是一次运行的像素纪念品。

生成并非“表达了什么”，而是“允许什么发生、禁止什么继续”。作者性不过是两套约束在某一帧达成的暂时共识。像素只是潜空间的有损翻译，它不是“思想的图像”，而是规则在时间里留下的痕迹。

展览：

• 2025 被协商的图像，日常的生成，天津美术学院。
• 2025 被协商的图像，拉古纳国际艺术奖亚洲首展-上海站。