Reflection

Concept and Motivation

After Project 4, I became curious about what happens when language becomes uncertain, ambiguous, or structurally unstable. In most AI workflows, prompts function like instructions: the more specific the phrasing, the more predictable the result. But I wondered what would happen if I stopped treating prompts as directives and instead approached them as poetic material, something open to interpretation, drift, and creative misreading. Could a prompt act less like a command and more like a fragment of language that the machine must struggle to interpret? These questions became the foundation for this project, which explores how meaning shifts when language loses clarity and moves toward abstraction, contradiction, and fragmentation. I wanted to see whether ambiguity could serve as a generative engine, pushing the machine into weirder or more expressive territory than clarity could.

Human–Machine Workflow

My process began with an AI-generated image from MidJourney. I distorted that image and then fed the distorted version into ChatGPT, asking it to describe what it saw in one sentence, but it couldn’t use literal or technical language, and it had to respond as if it were a poetic visual interpreter. ChatGPT’s output then became a new prompt for MidJourney. I would examine the resulting image, return to ChatGPT, and iterate by adding constraints, removing syntax, introducing contradictions, or asking for a more fragmented reading. Each description was sent back into MidJourney, and each resulting image became another invitation to reinterpret.

Working this way created a rhythm of alternating contributions. The model offered interpretations I wouldn’t have supplied on my own, and those interpretations helped me notice new directions for the next iteration. I adjusted prompts not to correct the system but to explore how different linguistic pressures like shortness, looseness, instability would reshape what the AI understood. The machine’s outputs served as cues for the next move, and each phase carried forward traces of the last.

Over time, the workflow felt less like issuing commands and more like maintaining a dialogue in which both sides influenced the evolving shape of the project. The boundaries between describing, generating, and reinterpreting became fluid, and the piece grew out of that ongoing exchange rather than from any single decisive action.

How My Thinking Evolved

Throughout the project, I noticed that the most interesting and least “sloppy” results didn’t emerge from descriptive, well-formed prompts. Instead, they came from prompts that were ambiguous, fragmented, contradictory, or structurally broken. I initially assumed clarity would give me more control over the output, like using a well-constructed sentence, a stable structure, a clear aesthetic intention. But the more I experimented, the more I saw the opposite happening.

This changed my understanding of what a prompt is and what prompting can do. I started the project thinking about prompts as tools for control, and I came away believing prompts could be engines of unpredictability and experimentation. The process taught me that “accuracy” in prompting can sometimes limit creativity, while uncertainty can expand it. My thinking evolved from trying to refine instructions to trying to destabilize them in intersting and productive ways.

Prompt Aesthetics

As the project developed, the prompt itself became one of the main creative materials I worked with. The wording, structure, and degree of ambiguity in each prompt shaped the direction of the images in ways that felt immediate and noticeable. Small shifts like removing syntax, introducing contradiction, shortening a phrase, or breaking a sentence apart often led to significant changes in the resulting visuals. This made the prompt feel less like a technical instruction and more like a medium with its own expressive possibilities.

Working this way made me more aware of how language operates inside generative systems. A prompt can hold tone, rhythm, and implication, and the AI responds to those qualities as much as it does to concrete terms. The images reflected not only what the words described but also how they were arranged, how open or closed they felt, and how much interpretive space they allowed. Over time, I started treating each prompt as an experiment in shaping that space—adjusting clarity, density, and structure to see how the system would interpret the shifts. My iterations show how small changes in language can change the direction of the output.

Creativity, Authorship, and AI

Throughout the project, my role shifted into a form of guiding, selecting, and shaping. I prompted the system, altered images, reintroduced them into the workflow, and responded to whatever the AI produced next. Each step created new material for me to consider, and the process became less about executing a fixed idea and more about staying attentive to what emerged.

As I continued working, I began to understand the AI’s contributions as part of the creative process. Its descriptions and interpretations often highlighted details I might not have focused on, or generated imagery that suggested directions I hadn’t anticipated. These moments gave the project a sense of momentum that came from the interaction itself. Rather than relying on a single vision, the work grew out of a sequence of exchanges—each prompt, description, and image providing something to build from.

This project did not come from one-sided control or from handing decisions over to the machine. It developed through a shared rhythm: I set constraints, the system produced variations, and the resulting material informed my next move. The creativity emerged through this ongoing process of responding and refining. Working with AI in this way expanded the range of possibilities I could work with and encouraged a more exploratory attitude toward making images and language.

AI Tools Used and Chat Link

ChatGPT (Free version)
MidJourney (Basic plan)

ChatGPT Prompt Iteration Chat Link: https://chatgpt.com/share/692d3a8a-7900-8005-9397-44032038a989