曹澍，《扩散》，2024，3D 数字 渲染影像，7’09’’，尺寸可变 

Cao Shu, Diffusion, 2024, 3D digital rendering moving image, 7’09’’, dimensions variable

作品《扩散》围绕幽灵学媒介展开叙事，探讨了摄影术、 AI 扩散模型技术与死者之间的复杂关系，以及核问题 带来的历史创伤。在驻留期间，艺术家研究了上海灵学 会和上海盛德坛的资料，其中包含关于降乩摄影的描述。 在艺术家看来，这些描述与时下热议的 AIGC 扩散模型 所处的技术情境有着惊人的相似之处。早期 AI 经常生 成多手指和四肢倒错的人类图像，其扩散模型生成的图 像仿佛人类的噩梦，充满了不合常理之处，唤起人们难 以名状的情绪。这也引发了一种都市传说——AI 大模 型的黑箱中寄宿着逝去之人的怨怒。 

作品尝试从早期 AI 的这一图像特质出发，将其作为叙 事的起点。昏暗的房间内，一位喃喃自语的摄影师困于 长崎核辐射的噩梦之中，误将 AI 生成的图像当成了噩 梦中所见的场景。而比基尼环礁遭受辐射的河豚，将自 己死后的影像永远留在了底片上。摄影如同一种幽灵的 痕迹，试图萦绕并提醒生者时间的意义 ......

Diffusion revolves around the narrative of ghostly media, exploring the entangled relationship between photography, AI diffusion models, and the deceased, as well as the historical trauma brought by nuclear issues. During the artist’s residency, he researched materials from the Shanghai Spirit Society and the Shengde Hall in Shanghai, which include descriptions of “ji spirit” photography. To the artist, these descriptions bear similarities to the technical context of the currently hotly debated AIGC (AI-Generated Content) diffusion models. Early AI often generated human images with extra fingers and distorted limbs, resembling human nightmares filled with irrational details and evoking indescribable emotions. This also gave rise to an urban legend: that the black box of large AI models harbors the resentment of the deceased.

The work takes this characteristic of early AI imagery as the starting point of the narrative. In a dimly lit room, a mumbling photographer is trapped in a nightmare of Nagasaki’s nuclear radiation, mistaking AI-generated images for scenes from his dreams. Meanwhile,

a radioactive pufferfish from Bikini Atoll leaves a permanent posthumous image on the film. Photography, like a ghostly trace, lingers and attempts to remind the living of the meaning of time ...