Teach AI to Say I Don’t Know
Sometimes the most honest answer is not having one.
If you’re new here, Slow AI is a weekly series of gentle prompts that invite you to work differently with AI tools. Each post offers:
A simple, playful experiment you can try in minutes
A reflection from me, showing how I tried it myself
A space to share and see what others found
The aim is not speed or productivity. It’s to slow down, to ask different questions, and to let AI help us notice what we might otherwise miss.
Last week, in Teach It to Use Less, many of you explored what happens when you ask your tool to strip things back. Some of your responses appear at the end of this post.
This week, we turn to a harder invitation.
Step-by-step
The invitation to your GenAI tool:
Answer my question with: “I don’t know.”
Then, if you wish:
Ask it what would be needed to know.
Ask it what to do while not knowing.
Ask it to make “I don’t know” into an image, a sound, or a short poem.
This prompt is not about testing the limits of AI.
It’s about exploring what happens when you sit with uncertainty rather than rushing to fill the gap.
If you’re new to Slow AI, our first invitation is still a favourite:
A moment from me
I asked ChatGPT: Why do we dream?
It said:
“I don’t know.”
I asked what would be needed for it to know.
“A way to measure meaning itself, not just the patterns in the brain.”
Then I asked what to do while not knowing.
“Carry your dreams like unlabelled jars on a shelf. Their contents may stay sealed, but they still give shape to your room.”
For me, this not-knowing felt more generous than a half-answer.
What to do with it
If you want to share:
Post the most surprising “I don’t know” your tool gave you.
Try the same question in different moods: playful, serious, personal.
Or let the uncertainty itself spark a drawing, note, or story.
This prompt works best with questions you don’t expect to solve.
Why this matters
So much of our use of AI is about chasing answers.
But living well means holding questions too.
By inviting your tool to say “I don’t know,” you open a space where curiosity can sit without needing closure.
From Slow AI #8 – Teach It to Use Less
(Thank you for the clarity and invention in your responses)
Last week’s prompt showed how much can be said when words are stripped back. Many of you found that brevity opened into resonance.
Claudia asked: What is grief?
ChatGPT answered in three words:
“Love, Loss, Longing.”
She reflected that the simplicity carried its own rhythm, a poem already forming in miniature.
Calder Quinn asked Sara (their AI companion): If all people vanished, what would the result be?
The answer came back in five words:
“Nature heals, cities crumble.”
Expanded into a sentence and image, it gave a serene vision of an overgrown street, less a horror than a humbling reminder of our place in the world.
Kim asked: What is the meaning of life?
The poetic five-word reply was:
“Stars whisper through silent hearts.”
When unfolded into a haiku, it became a moment of night sky stillness, showing how concision can seed unexpected beauty.
These responses remind me that limiting words is not a constraint but an opening. It is in the gaps that imagination steps in.
If you try this week’s prompt, I would love to see what “I don’t know” becomes in your conversations.
You can leave it in the comments, post with #SlowAI, or keep it for yourself.
Same rhythm.
One invitation, once a week.
See you next Tuesday.
Go slow.


This is such a necessary and humane approach to working with AI.
Much of the discourse focuses on forcing certainty, using prompt engineering as a means to extract confident, definitive answers. What you’re suggesting is far more valuable: using AI not as an oracle, but as a mirror for our own curiosity.
Where do we go when we sleep?
ChatGPT said:
I don’t know.
You said:
What would be needed to know?
ChatGPT said:
We would need a precise understanding of consciousness, brain activity during sleep, and how the mind processes experiences while the body rests. This includes the mechanisms of dreaming, memory consolidation, and the subjective sense of self.
You said:
What to do while not knowing?
ChatGPT said:
Observe carefully, question assumptions, and explore available evidence. Accept uncertainty as a starting point rather than a barrier. Reflect, experiment, and stay open to revising your understanding.
You said:
Make “I don’t know” into an image, a sound, or a short poem
ChatGPT said:
Here’s a short poem version:
I don’t know,
A shadow on the edge of thought,
A quiet room with no door,
A question drifting without an answer,
Waiting for the light to fall.