At a recent Australian Computing Society panel, I found myself grappling with a familiar question from the audience: "Can people truly be creative when AI can do all the 'hard' work, especially with image generators creating visuals from a couple of words?”

The concern isn't surprising, really. The narrative around AI innovations seems stuck in a loop of productivity metrics and efficiency gains. It’s all business. Just do your work faster so you can…do more work? It's no wonder people are experiencing AI fatigue, leading to a surge of startups promising to put control back into creative’s hands.

However, I've come to firmly believe that AI can enhance and expand our creative process rather than replace it. The key? Approaching these tools as entirely new mediums for exploration, rather than a magical end-all-be-all.

Bringing creative mess to AI

I'm no stranger to image generators, but my approach isn't typical. The promise of efficiency leads many to approach these models with a specific end goal in mind—a particular image for a project or a visual component of a larger work. It's a valid use case, certainly. The expectation is that these models will produce whatever is in your mind, or at least close to it, if only you can describe it exactly.

We're now realising that describing it exactly doesn't always get you what you want, so the makers of these models are bringing in improvements like style references, image uploading, and prompt expansion to help get you there. These are amazing tools, as pictures are indeed worth a thousand words.

However, I find this a losing game. The more you try to control an output, the more it gets away from the vision in your head.

I prefer to expect nothing now. This approach evolved quite accidentally while testing different models as part of my role. Bouncing between Midjourney, Stable Diffusion, FLUX, and DALL-E, I wanted to push it to the limits, but how could I do that in the structure of test prompts? Instead of pursuing perfect outputs, I began prompting abstract concepts: a feeling of longing, the weight of unspoken words, or a literary quote capturing the essence of longing. The first test was an incredibly nostalgic moment for me, a quote from Plato’s Allegory of the Cave.

“Therefore, of this matter itself, there must be a craft of some kind, which would be a most efficient and effective means of transforming the soul. It would not be an art that gives the soul vision, but a craft at labor under the assumption that the soul has its own innate vision, but does not apply it properly. There must be some kind of means for bringing this about.”

The results were extraordinary - not because they were perfect, but because they were unexpected. Suddenly I was getting outputs I would have never gotten to without that play. It was expanding my imagination of what I could do, where I could refine a concept further. Now, it's the process I use in my creative exploration, a bit of a conversation with my imagination.

I'd like to share this approach with you–not as a formula for success, but as an invitation to explore. Perhaps you'll discover your own unique way of using these tools or find some unexpected moments during your creative process.

The Scrapbook and the Dark Room

My process with AI involves two conceptual spaces:

The Scrapbook: This is where I collect and combine ideas—notes, style references, quotes, and descriptions. It is messy and it’s meant to be. It’s scattered across my brain and Notion.
The Dark Room: Here, I experiment with the AI. I trial and error, push and pull, constantly referring back to my scrapbook and imagination to bring my vision to life. This is done in a Canva whiteboard.

This process has just enough structure to get me started and collect my thoughts in one area.

Illustrating the process

Let me walk you through a recent project inspired by The Tale of Genji. My scrapbook for this project was a digital mess of seemingly unrelated elements:

Literary quotes that produced unexpected results in Midjourney
Style reference codes from the active Midjourney community. Combining them was so much fun to see what happened.
Experiments with personalisation codes. At the time, it was a new feature in Midjourney and I wanted to see how it worked.
Notes from my recent classical literature reading. It’s a respite from reading current affairs and tech news.

The scrapbook should be messy. Don’t try to bring too much order into it. Bringing the disparate elements together helps coalesce them into something new and unexpected.

In the darkroom, everything comes together on my whiteboard. Like a detective's evidence board, it helps me see connections I might have missed. Each generation gets documented with its prompt and parameters, along with personal notes for future reference. It's a process of refinement, but not in the traditional sense - more like following breadcrumbs of inspiration.

**Finding flow**
The beauty of this approach is how it combines structure and chaos. The scrapbook phase lets your mind roam free, collecting inspiration like a magpie collecting shiny objects. The darkroom phase gives you space to experiment with these findings, while keeping track of your journey. For example, without adding specific parameters around the Japanese Heian period, when the Tales of Genji was written, all the outputs were considerably Western. I wanted to convey more of the atmosphere of Murasaki Shikibu’s writing.

Remember: this isn't about reaching a perfect destination. It's about enjoying the process of discovery, letting each experiment inform the next, and staying open to the unexpected magic that happens when you give creativity room to breathe.

Enjoying Imperfect Creation

Returning to that question from the ACS panel–can we truly be creative in an AI-augmented world? I've come to realise we're asking the wrong question entirely. Like the shadows in Plato's cave, we're fixating on the outputs rather than understanding the deeper transformation happening in our creative process.

What I've discovered through this messy, imperfect journey is that AI isn't replacing creativity–at least for me, it's revealing new dimensions of it. Each unexpected output, each "failed" experiment becomes a mirror, reflecting back aspects of my creative taste I hadn't fully recognised. Why am I drawn to certain outputs? What makes some generations sing while others fall flat? The AI isn't just generating images; it's helping me understand my own aesthetic instincts on a deeper level.

Here's the truth that took me months to accept: not everything needs optimisation. Through my own stumbles and discoveries, I've found that success lies in treating AI as a creative collaborator rather than a productivity tool. Start with that messy scrapbook of ideas–let it overflow with seemingly unrelated concepts. Give yourself permission to explore dead ends, to spend hours following creative threads that might not lead anywhere immediately useful.

The future belongs to those who can embrace uncertainty and find inspiration in the unexpected. So start your own scrapbook, set up your digital darkroom, and begin your journey of discovery. Perhaps the most revolutionary act isn't maximizing productivity; instead, it's reclaiming the joy of creative exploration for its own sake, all while learning more about yourself as a creator in the process.