When AI Surprises You: The Joy of Unexpected Results

I’ve started to realise that the most interesting moments with AI aren’t the ones that go exactly to plan. They’re the ones that don’t.

You type in what you think is a precise prompt, expecting a predictable outcome and the machine goes completely off script. Sometimes it misses the point spectacularly. Other times, it creates something that feels oddly beautiful.

That’s partly because Large Language Models (LLM’s) like ChatGPT don’t actually understand what they write, they predict the next most likely word or image based on patterns they’ve learned from billions of examples. Every so often, those predictions take a creative detour, and that’s when things get interesting.

And it’s in those unexpected moments that I find the most joy. They remind me that creativity, whether human or artificial, isn’t about control. It’s about discovery.

When AI produces a strange image, a misplaced metaphor, or a curious turn of phrase, our brains light up trying to make sense of it. The AI isn’t being clever, we are, as we interpret and shape its chaos into meaning.

The Creative Value of Mistakes

The joy of experimenting with AI is that you are never sure what response you will get from your prompt. The other joy is that if you put the same prompt in again you are likely to get a completely different response.

I’ve been experimenting with putting some first lines from famous poems and novels into Midjourney to see what sort of image I get back. This is one interpretation it had of William Wordsworth’s, “I wandered lonely as a cloud.”

This is what it made of, “Once upon a time and a very good time it was there was a moocow coming down along the road and this moocow that was coming down along the road met a nicens little boy named baby tuckoo…” The beginning of ‘A Portrait of the Artist of Young Man’ by James Joyce.

These four images are a selection of what Midjourney gave me when the prompt was, “Twas brillig, and the slithy toves / Did gyre and gimble in the wabe:” The beginning of the nonsense poem Jabberwocky by Lewis Carroll.

We’ve always celebrated “happy accidents” in art, the paint splatter that adds character, the typo that suggests a new phrase, the unexpected chord that becomes a song. AI mistakes are just the digital version of that.

Maybe the future of creativity isn’t about preventing mistakes, but making friends with them.

The Prompt That Went Wrong (and Right)

I have experimented with purposely making a mistake with a prompt, eg. instead of saying to AI Claude, “Write a 150 word monologue in the voice of a boy with green hair.” The prompt actually read, “Write a 150 word monologue in the voice of a buoy with green hair.”

This was the result:

---

Adrift In Vanity

Bobbing here day after day, watching the same patch of sea. The gulls laugh at my seaweed mane, they don’t understand style. “It’s GREEN!” I shout to passing boats, but they never listen. Just flash their lights and move along.

The tide pulls at me, wanting to take me dancing, but my chain keeps me loyal to this spot. Sometimes I dream of breaking free, riding massive waves to distant shores where other buoys might appreciate my verdant locks.

Winter storms are the worst, they tangle my beautiful hair something fierce. Summer brings tourists who point and say, “Look at that weird buoy!” I’m not weird, I’m DISTINGUISHED.

At night, when the moon reflects on the water, I see my silhouette and think, “You magnificent buoy, you.” Not everyone can pull off green hair in this vast blue expanse. I’m a maritime trendsetter, darling. A lonely, anchored trendsetter.

---

Sometimes the best thing an AI can do is surprise you into writing differently.

Learning to Embrace the Unexpected

Working with AI has changed the way I think about control. We tend to see these tools as obedient assistants,  precise, reliable, efficient. But in creativity, obedience can be boring. The best results often come from a gentle misunderstanding or a wild leap of logic.

Now, when I get an odd or offbeat result, I pause before editing. I ask what it’s trying to tell me. Could this “mistake” contain a more interesting idea? What happens if I run with it instead of correcting it?

The results aren’t always good, but they’re rarely dull.

From Chaos to Curiosity

This feels like a natural continuation of my earlier post, When Chaos Tells a Story. Then, I was testing whether AI could find meaning in pure randomness. Now, I’m discovering that meaning often begins in randomness.

The joy isn’t in the perfect prompt or the flawless output. It’s in the tiny surprises that make you stop and think,  the moments when neither human nor machine is fully in control. That’s where imagination lives.

A Closing Thought

AI doesn’t feel surprise, but it can still cause it. And that’s a powerful thing.
When an AI result makes me pause, smile, or rethink an idea, it’s doing what art has always done, making me see the world differently.

Maybe surprise is the truest sign that AI has entered the creative process, not because it’s thinking, but because it’s helping us think differently.

When that happens, the question isn’t “Did the AI understand me?”
It’s “What have I just discovered about myself?”

---