I apologize for the delay in completing this, as I was without access to a computer for two days. Here we go…
In this project, I decided to explore the realms of AI creation through the lens of semiotic structuralism. My goal was to understand where AI surpasses the boundaries of linguistic theory and unravels what Jacques Derrida might describe as instances of "Différance." Derrida's work revolves around the concept of Différance, which highlights the intricate relationship between words and the meanings they signify.
Différance suggests that words and signs are not direct equivalents of the concepts they represent. Instead, meaning emerges from the distinctions they hold in relation to other words and signs. This constant differentiation leads to the creation of binary oppositions and hierarchies, resulting in meaning that perpetually defers, continuously delayed through a series of signifiers. Derrida uses terms like "espacement" (spacing) and "temporisation" (temporizing) to illustrate this process. In essence, with every articulation of a signifier, the signified both "differs" and "defers" in meaning. Derrida coined a new term for this phenomenon, known in post-structuralist theory as "Différance."
The project led me to contemplate Derrida's ideas, particularly in relation to the generation of entirely new sets of images with minimal or sometimes no changes to the original prompt. The "signification" in image generation varied significantly with each new "signifier" (prompt). I attempted to mitigate this variation using Midjourney's "--seed" parameter, which is designed to produce consistent results for the same prompt and seed combination. However, I was still astounded by the diversity of outcomes.
Let's delve into the specific images generated:
Image one
Prompt: Van Gogh's painting The Starry Night in the style of Ansel Adams, seen through a car window, using high contrast black and white Illhord HP5 with cinematic lighting, photorealistic --aspect 16:9 --chaos 5 --seed 14 --style raw --s 250 - @MAT 255 (fast)
From the very beginning, I needed to consider the two crucial elements of the prompt: Van Gogh's renowned painting "Starry Night" and the requirement to depict it through a car window. This is where the project became intriguing. Two of the images from the first set, displayed in our Discord channel, portrayed the painting through the car's windshield rather than through the windows. The proportions of the car's interior were also distorted, perhaps due to the absence of specifications regarding the camera lens in the prompt. In terms of aesthetic style, although the color tones were high-contrast black and white, they did not quite resemble Ilford HP5. Instead, they evoked the feel of a linocut print with black ink on paper. Nonetheless, all four images, despite their distinctions, shared commonalities. They depicted a view from the perspective of someone inside a car and exhibited similar tonal and color qualities.
However, what didn't align with my expectations was the lack of photorealism. The images, though compelling in their own right, didn't meet the photorealistic standard set by my prompt.
Image two
Prompt: Van Gogh's painting The Starry Night in the style of Studio Ghibli Japanese anime, seen through a car window, with vibrant colours:: --aspect 16:9 --chaos 5 --seed 14 --style raw --s 250 - @MAT 255 (fast)
Continuing with the same prompt for the image elements but varying the aesthetic parameters, I revisited "Starry Night." However, I wasn't sure if the color tone, illustration, and contrast resembled the Studio Ghibli style or veered closer to Isao Takahata's work. Distortions in the proportions of the car interior were evident, and in some cases, it was challenging to distinguish the car's interior from its exterior. Despite these variations, there was a common thread—they all depicted scenes from inside or near a car and shared similar color and tonal qualities.
In these images, Van Gogh's painting was more discernible, which I had anticipated to some degree. Even the second image, unexpectedly resembling a more photorealistic representation, retained the color tones of the original painting. However, it deviated from the anticipated anime style.
Image three
Prompt: Van Gogh's painting The Starry Night in the style of line drawing, seen through a car window, use pencil sketching::3 --aspect 16:9 --chaos 5 --seed 14 --style raw --s 250 - @MAT 255 (fast)
With the same prompt and varying aesthetic parameters once more, the image I chose displayed a line drawing of a car's interior. The others leaned toward photorealistic 3D animation or video game aesthetics. The aesthetic styles among these three images diverged significantly, but I selected the one that adhered most closely to the prompt. However, it surprisingly omitted the central element: Van Gogh's painting. Instead, the image depicted a car that appeared to have been in a crash, with a shattered windshield.
This set of images deviated substantially from the prompt, and only the line drawing aesthetics of the third image aligned with my expectations. Nevertheless, the patterns of the windshield cracks seemed to echo the turbulent sky in Van Gogh's painting, which was an unexpected and pleasant surprise.
As far as the prompt was concerned, these images largely missed the mark, yet they provided a unique interpretation of Van Gogh's work in a different medium and setting, making them thought-provoking.
Image four
Prompt: Van Gogh's painting The Starry Night in the style of a pencil sketch, seen through a car window::3 paper::3 --aspect 16:9 --chaos 5 --seed 14 --no human --style raw --s 250 - @MAT 255 (fast)
In this instance, I expected the AI to produce hand-drawn pencil sketches of Van Gogh's "Starry Night." However, my use of the "::3 paper" query emphasized the need for the images to resemble paper cutouts. As I delved into this project, I had to rely on the free Midjourney Prompt Helper to refine my prompts, ensure consistent parameters across the board, and gain a better understanding of prompt technicalities.
Throughout this experiment with Midjourney, I found it necessary to utilize the free Midjourney Prompt Helper (available at
https://promptfolder.com/midjourney-prompt-helper/). This tool proved invaluable in refining the linguistic parameters of my prompts, ensuring precision and conciseness. It enhanced the determinism of the results, allowed me to maintain specific parameters across multiple images (facilitating the analysis of subtle differences in images generated from nearly identical prompts), and provided guidance on the technical aspects of using prompts effectively.
The "::" command was particularly intriguing, as it prioritized subsequent prompts. In this case, "::3 paper" emphasized the use of paper or paper cut as the formal aesthetic element. The images shared some commonalities—images 1 and 4 appeared as paper cut versions of "Starry Night," while images 2 and 3 didn't capture the turbulent sky. The unexpected absence of the car in images 1 and 4 underscored the importance of precise phrasing in the prompt. While images 2 and 3 fulfilled expectations by displaying car windows, the aesthetic style leaned toward photorealism rather than resembling pencil sketches. In essence, the images worked well in generating unforeseen aesthetic results, even if they didn't strictly adhere to the prompt.
Image five
Prompt: Van Gogh's painting The Starry Night in the style of a 3d unity render, Octance render, 3d sculpture, raytrace, 4K, focused, detailed, seen through a car window::3 --aspect 16:9 --chaos 5 --seed 14 --no human --style raw --s 250 - @MAT 255 (fast)
This set of images closely aligned with the prompt. The AI excelled in rendering an Octance render/3D Unity render look, featuring car windows and interpretations of Van Gogh's turbulent sky. However, the absence of randomness in these images, contrary to my prior experiences with Midjourney, was an unexpected outcome.
While they adhered to the prompt, the absence of randomness was a slight departure from my expectations.
Image six
Prompt: Van Gogh's painting The Starry Night in the style of plasticine clay modelling sculpture, seen through a car window, use bright and vibrant colours:: --aspect 16:9 --chaos 5 --seed 14 --no human --style raw --s 250 - @MAT 255 (fast)
Here, I tasked Midjourney with rendering a Plasticine clay modeling aesthetic. The resulting images evoked the style of stop-motion clay animation films, but none adhered to the prompt's requirement of showing scenes through car windows. Instead, they depicted the entire cars.
This unexpected divergence from the prompt was precisely what I was looking for in my previous experiment, where I sought playful imagination rather than strict adherence. It was a reminder that AI, when allowed some creative freedom, can introduce an element of unpredictability and creativity into the process.
Image seven
Prompt: Van Gogh's painting The Starry Night in the style of lino-cut print, seen through a car window, use ink:: --aspect 16:9 --chaos 5 --seed 14 --no human --style raw --s 250 - @MAT 255 (fast)
The prompt was to generate the same image in the style of linoleum print. While it's arguable whether the AI's output could be classified as linoleum prints, the results were striking in their reinterpretation of Van Gogh's work through different aesthetic parameters. The images displayed randomness in the choice of pictorial elements, as the car windows appeared in just three of them only when I zoomed out of the images that were originally generated (Zoom out 2X). Yet, there was a thematic cohesion across all the images, especially in their emphasis on blue colors and post-impressionist brush strokes.
Transposing a medium often necessitates a shift in style, and transliteration demands creative originality. Midjourney showcased this creativity through its random interpretations. While the images deviated from the prompt, they managed to retain the essence of Van Gogh's work and presented an enjoyable departure from my stipulated expectations.
Image eight
Prompt: Van Gogh's painting The Starry Night in the style of a monochromatic charcoal sketch on paper, seen through a car window, :: --aspect 16:9 --chaos 5 --seed 14 --no colour --style raw --s 250 - @MAT 255 (fast)
This set of images resembled the first set, with visible charcoal strokes and a lack of photorealism. They maintained a consistent aesthetic quality of charcoal sketch on paper and displayed car windows or windshields in the scenes. To some degree the brush strokes resemble that of image seven. I was eager to see how AI would translate the vibrant and colorful "Starry Night" into stark monochromatic charcoal strokes on white paper. The results proved unexpectedly wonderful, capturing the essence of Van Gogh's turbulent sky and brushwork in an entirely different medium.
The images worked exceptionally well because they adhered to the prompt, transforming Van Gogh's oil-on-canvas style into a virtual charcoal rendition.
Image nine
Prompt: Van Gogh's painting The Starry Night in the style of a charcoal sketch on paper, seen through a car window, :: --aspect 16:9 --chaos 5 --seed 14 --no human --style raw --s 250 - @MAT 255 (fast)
In this case, I introduced a ninth image where Midjourney, instead of strictly following the prompt, embraced creativity. In this case, the AI didn't follow the prompt closely, creating something random. This randomness is an essential aspect of AI's creative process, offering new and unexpected perspectives. Often, when testing new technology, we focus on how closely it can adhere to human expectations. But I think that it is essential to recognise AI's unique abilities instead of expecting it to mimic human expectations verbatim. Each image generated by the AI exhibited various levels of conformity to the prompt, and the unexpected variations often produced intriguing results. The AI's creative approach allowed for new interpretations of Van Gogh's masterpiece, adding depth to the project's exploration of AI's capabilities.
We tend to overlook the fact that machines, including AI, occupy a unique space between and serve as intermediaries between human labor and the work produced. They can reveal a new and unexpected world, going beyond being mere tools for convenience. Martin Heidegger's concepts of “Techné” and “Aletheia” [in Heidegger, Martin. (1977). The Question Concerning Technology] emphasise the idea that there is nothing inherently technological about technology. AI engines like Midjourney can provide surprising and creative results that enrich our understanding of the world.
