This as-told-to essay is based on a conversation with Boo Kok Chuon, the chief operating officer of a law firm in Singapore. The conversation has been edited for length and clarity.
Since early last year, we took a bold step in introducing ChatGPT to Kiki, our 8-year-old daughter.
We thought that we should start our kid early, given how technology is progressing these days.
She has been frequently communicating with her own ChatGPT. Whenever she faces an issue, she will go to ChatGPT to ask for a solution.
Once she spilled a drink on a shirt and she felt bad. She asked ChatGPT how to resolve it. She also asked ChatGPT how to take care of her turtles.
Then we tried something bigger.
Vibe coding an app together
We built a companion app for our physical game Mu Jong, a music-learning mahjong-style game. We did it within three hours.
We used a combination of generative AI tools, including ChatGPT for prompt development and logic structuring, Base 44 for interface development, and Nano Banana Pro for visual assets.
The app has a score tracker, a keyboard music feature, and a countdown timer. Kiki conceptualized the features and guided the process, while I supported her in structuring prompts and translating her ideas into instructions for the AI tools.
We created a mascot for our app. Kiki drew out the mascot on a piece of paper. It was quite a crappy drawing, but once you upload the photo into the AI, it will beautify it. She tried it, and was like: ‘”Wow, it’s like magic.” Suddenly, her drawing looked commercially ready.
I also guided her. I told her: “You have to say the purpose of the drawing, and what you want to use this mascot for.” We refined the drawing based on further prompting the AI.
Children should be exposed to AI early
It is very important for kids — especially like Kiki — to be exposed to AI. Of course, human input is still required for review.
She was born into a privileged generation in which technology was already very advanced. Nowadays, kids may not be so easily impressed, given how readily available all this technology is.
Kiki is learning to break a problem down into steps and use AI tools to test and refine her ideas. She also learned how to use natural language to construct more structured prompts. For example, she began specifying attributes such as “rounder shape, softer colors, friendly expression,” which helped her achieve more consistent outputs.
For her devices, we have oversight over the content she engages with.
On YouTube, we constantly review what she watches and look at her internet history.
We are very open with her and want to be transparent and respectful. She’s aware that we are supervising her.
At any point in time, if we see anything that is not quite right, the approach that we use is to guide her and not judge her. We will ask her, “Why do you do this? Do you do this because it helps you with certain issues that you are facing?”
And from there, we will slowly — again through this questioning method — lead her toward the more appropriate actions.
There are some inherent issues with AI. Sometimes, AI can be a bit too conservative in its outputs.
I think that if you rely on AI too much, it might be emotionally dangerous in a way, too. We should not trust these systems 100%, and take a cautious, objective approach when dealing with them.
Do you have a story to share about coding with AI? Contact this reporter at cmlee@businessinsider.com.
