In this paper, we argue that Midjourney—a generative AI program that transforms text prompts into images—should be understood not as an agent or a tool, but as a new type of artistic medium. We first examine the view of Midjourney as an agent, considering whether it could be seen as an artist or co-author. This perspective proves unsatisfactory, as Midjourney lacks intentionality and mental states. We then explore the notion of Midjourney as a tool, highlighting its unpredictability and the limited fine-grained control users have over its outputs, which sets it apart from traditional artistic tools. We propose that creating images with Midjourney parallels a gardener's interaction with natura naturans—nature's autonomous generative processes. Similarly, Midjourney acts as machina naturans, where users engage with its dynamic recalcitrance to influence, but not fully control, the creative process. This approach presents Midjourney as an artistic medium shaped by its inherent unpredictability.

中文翻译：

---

塑造形象：生成式人工智能和园艺媒介


在本文中，我们认为 Midjourney（一种将文本提示转换为图像的生成人工智能程序）不应被理解为代理或工具，而应被理解为一种新型的艺术媒介。我们首先审视中途作为代理人的观点，考虑它是否可以被视为艺术家或合著者。事实证明，这种观点并不令人满意，因为《中途》缺乏意向性和精神状态。然后，我们探讨了 Midjourney 作为工具的概念，强调了它的不可预测性以及用户对其输出的有限细粒度控制，这使其与传统艺术工具区分开来。我们建议，用中途创作图像类似于园丁与自然的互动——自然的自主生成过程。类似地，《中途旅程》扮演着机械自然者的角色，用户以其动态的顽抗性来影响但不完全控制创作过程。这种方法将中途呈现为一种因其固有的不可预测性而形成的艺术媒介。