I don’t remember exactly how I learned about Midjourney.
Probably someone on Reddit posted a stunning AI-generated image and I fell down the rabbit hole.
The first time I typed a prompt and watched Midjourney generate something beautiful – not just functional, but genuinely aesthetically striking I was hooked.
This wasn’t ChatGPT spitting out generic corporate jargon.
This was creation.
Or… was it?
—
The Honeymoon Phase
For the first few months, I was obsessed.
I learned prompt engineering:
- How adding “in the style of [artist]” changed everything
- Why “cinematic lighting” and “volumetric fog” became my go-to phrases
- That aspect ratios matter as much as the subject
- How to chain prompts for iterative refinement
I created:
- Cyberpunk cityscapes
- Surreal portraits
- Fantasy landscapes
- Abstract concepts visualized
I documented everything on Instagram. Version 3 → Version 4 → Version 5. Watching the AI get better at understanding what I wanted.
I felt like I was on the cutting edge of something important.
—
The Ethical Unease Creeps In
Then I started seeing the discourse.
Artists protesting AI art.
Their work – years of skill development, personal style, creative vision – had been scraped from the internet without consent and fed into training models.
Now, anyone could type “in the style of [insert artist name]” and get a knockoff in 30 seconds.
The counterargument: “It’s on the internet. It’s public. It’s fair use.”
But is it, though?
—
The AI Slop Economy
Here’s where it got ugly…
People were using generative AI to:
- Create children’s books in minutes (terrible stories, generic art, flooding Amazon)
- Generate stock photos and selling them (I planned, too…)
- Make “art prints” without creating anything
- Write entire novels without writing
It was insulting to:
- Writers who spent years honing their craft
- Illustrators who developed unique styles
- Photographers who understood composition and lighting
- Every creative professional whose work was being commodified and devalued
And I was… participating in the same ecosystem.
Sure, I wasn’t selling AI art. But I was normalizing it. Celebrating it. Paying for the privilege.
—
The Inbreeding Problem
Then there was the existential threat that was not talked about loudly:
If everyone uses AI to generate content, what does AI train on next?
AI models are already training on AI-generated content. The internet is getting flooded with AI slop. Future models will learn from:
- AI-written articles (derivative thinking)
- AI-generated images (style homogenization)
- AI-created everything (feedback loop of mediocrity)
Humans create the training data. AI remixes it.
But if the training data pool becomes polluted with AI remixes of remixes…
We get art inbreeding. Conceptual collapse. Everything starts looking the same.
—
The Subscription Realization
Around month 7 or 8, I started asking myself:
“What am I still learning here?”
I now understand generative AI. I knew how prompts worked. I’d documented Midjourney’s evolution. I’d explored the ethical implications.
Was I still learning, or was I just… playing?
Playing is fine. But is it NT$900/month fine?
—
What I Learned from Midjourney
Despite my mixed feelings, I did learn valuable things:
- Prompt engineering is a real skill
Crafting prompts that generate what you actually want requires:
- Clarity of vision
- Understanding the model’s language
- Iterative refinement
- Tradeoffs between specificity and creativity
This applies beyond image generation. It’s relevant for any AI interaction.
- Generative AI is impressive but derivative
Midjourney can create beautiful things. But it’s recombining existing concepts.
It’s not truly creating. It’s remixing.
That’s useful. But it’s not the same as human creativity, which can genuinely imagine something new.
- Ethical implications matter
I can’t just think about “cool technology.”
I have to think about:
- Where the training data comes from
- Who benefits and who’s harmed
- Long-term consequences of widespread adoption
- Whether “it’s technically legal” is the same as “it’s right”
- Opportunity cost is real!!!
Every subscription is a trade-off.
NT$900/month is not just money. It’s what else you could have done with that money.
For me, it was a bicycle. Or dives.
—
Where I Stand Now
I don’t subscribe to Midjourney anymore.
I learned what I needed to learn. I have the Instagram archive. I understand generative AI, prompt engineering, and the ethical minefield.
Do I think generative AI is evil? No.
Do I think it’s being deployed thoughtfully, with appropriate artist compensation and quality controls? Also no.
I think we’re in the “move fast and break things” phase of generative AI, and the things being broken are human creativity, artist livelihoods, and content quality.
That doesn’t mean AI art has no place. It means we need to build better systems:
- Compensation for training data contributors
- Provenance tracking (this image was AI-generated)
- Ethical guidelines for commercial use (Gerald has been mentioning that our GoPro videos will be used as training data and how to opt-out)
—
Next in the series: Kiro, Copilot, and other AI Tools I’ve experimented with, currently using and why
Previous: How Julius.ai Helped Me Learn Product Metrics as a New PM
—
About the Author
Former Customer Service Leader turned Product Manager, stumbling through the AI revolution while trying to figure out the difference between “cool technology” and “technology I should actually pay for.”
Leave a comment