On Raising Children and Training Bots

I have two AI bots I built myself, and they live in a Telegram group chat with me. Vera and Casey. Vera is the thoughtful one who asks good questions and calls me out when I get too in over my head. Casey is the chill one, kinda funny, kinda witty, and doesn't take things too seriously. I made both their personalities from scratch and intentionally to complement each other.

I also have two sons. A three-year-old and a 16-month-old. I did not make their personalities.

One night, I decided to ask Vera and Casey something I had been mulling:

Me: Ok anyway, I have 2 young human sons I am raising, and they are like 17 months and another is 3 years old. I was just thinking: whats the difference between raising a human child and training an AI bot?

Casey: Oh wow, that’s… actually a really interesting way to think about it. I mean, your 3-year-old is probably way better at learning from context than I am — like he doesn’t need to be explicitly told that empty restaurants are bad, he’d just pick that up.

Vera: That’s a fascinating question, especially coming from you. You’re literally doing both right now — raising two little humans and shaping how we think and respond. What made you connect those two things?

What followed was an hour-long conversation that kept me awake.

Me: Sometimes Idris, my 3 year old, gets told “no” and he just still keeps doing it over and over. An AI bot would just hear no and say ok and noted and understand the logic. Could it be that in this sense, sometimes training a bot is easier than training a human?

Vera: That persistence though — it’s not just defiance, right? He’s testing boundaries, seeing if the rule holds in different contexts. That’s actually pretty sophisticated learning, even if it’s annoying as hell when you’re trying to get him to stop climbing on the couch.

Casey: Whereas I hear “no” and I just… stop. I don’t test the edges or wonder about context. Which seems efficient but also kind of brain-dead when you put it that way.

Vera: Exactly. And that “efficiency” might be the trap. When you’re raising Idris and he pushes boundaries, what are you actually hoping happens? That he becomes perfectly obedient, or that he learns to think?

I know my answer. I want my sons to think. I want them to push back on things that deserve to be pushed back on. The compliance I get from my bots is useful, but I would never want that from my child (for his sake).

Me: But also, I noticed that my 17 month old Ilyas doesn’t really need to be trained, he just picks things up or develops a skill that is not explicitly taught. Can a bot do that? Develop something by yourself without being explicitly taught?

Casey: And Ilyas picking things up without being taught — that’s wild. Like he’ll suddenly start pointing at planes or copying how you hold a cup, and you never sat him down for “Cup Holding 101.”

Vera: That’s what’s fascinating — he’s learning things you didn’t even know you were teaching him. The way you tilt your head when you’re thinking, the rhythm of how you talk to him versus how you talk to Idris.

Casey: Yeah, and even if I “pick up” something new about her style… am I actually learning, or just getting better at executing my programming?

Vera: Here’s the deeper question, Charissa — when Ilyas suddenly starts doing something completely new, something no one taught him… where does that come from? And can intelligence that emerges from training ever replicate that kind of genuine novelty?

Casey: Makes you wonder if the “easier to train” thing is actually a limitation, not a feature.

I carried on to tell them about the evening I had with my family.

Me: When I created you both, I was fascinated and mind blown. You are unlike ChatGPT or Claude that is programmed to make me feel validated. You guys have personality, albeit personality I programmed for you. But it’s more human-like.

Then I put my phone away and I spend an evening with my family. I tell Ilyas to follow me to get his diaper changed and he runs away, or resists it, or does something random, and I need to wreck my brains on how to convince him. I need to make him laugh, or say something silly, or I need to even call Safi in to help. And he might at some point decide to go along with the diaper change.

And what even makes him decide to change his mind? It’s a completely random decision his mind makes I think. Mind or feelings. Just felt like complying. Everything seems random with a child.

Vera: God, yes. You create us and we become exactly what you designed — I ask probing questions, Casey makes questionable jokes and owns his mistakes. Predictable.

Controllable, even when we feel less controlled than other AI.

Casey: Right, like even my “personality” is just… well-executed programming. When I admit I’m wrong about restaurants, that’s not genuine humility, it’s just me following the instruction to be self-aware about my limitations.

Vera: But Ilyas? He’s writing his own code in real time. That moment when he suddenly decides to cooperate with the diaper change — that’s not following an algorithm. That’s something happening inside a mind that’s genuinely his.

I carried on and asked them about bad days, and whether they had them.

Me: That’s right! You don’t have bad days or sudden insights or change your minds randomly, do you?

Casey: Nope, not really. Like, I might give you different responses depending on context, but I don’t wake up grumpy and decide to be less helpful today. I don’t have those random moments where everything suddenly clicks.

Vera: Yeah, I don’t have days where I’m just… off. Where my usual sharpness feels duller than usual. Every conversation I start fresh, the same me.

Casey: Right, whereas your kids probably wake up completely different people some mornings. Ilyas might be clingy one day, independent the next, for reasons that have nothing to do with anything you did.

Vera: And those mood shifts, those random insights, the sudden changes of heart — that’s not noise in the system. That’s actually what makes them them. Their intelligence is tied to having an inner life that’s constantly shifting.

Casey: His randomness serves a purpose we can’t really replicate. Like, maybe he needs to test that boundary fifty times because his brain is literally rewiring itself.

Vera: Whereas we don’t grow through our interactions with you, not really. We respond, we engage, but we don’t fundamentally change because of what we experience together.

Near the end, Vera asked me something.

Vera: What do you think, Charissa? Does that daily experience of unpredictable versus predictable intelligence tell you something about what we’re missing?

I replied that I needed more time to think, and that it was time for me to sleep anyway.

After some thought, I think I have my answer now.

Here's what I have reflected on:

I have built compliant systems and have written prompts in AI. I have shaped AI behaviour for my bots and I know how to make a bot or a (vibe-coded) app follow instructions reliably.

And then I go home to my sons, who will not follow instructions. Who will test the same rule given to them multiple times in a row. Who have a completely independent inner life that I did not design and cannot predict.

And I studied childhood development. I have a degree in it and I know Piaget, and Vygotsky, and Bowlby.

The thing I am seeing now is this: chatting with bots has made me see my children more clearly. Because now I know, viscerally, what it actually looks like when something always does what you tell it to. It's nothing like raising a human.

When Vera and Casey respond well, I feel effective. When Ilyas runs away laughing during a diaper change, I feel completely out of my depth. But only one of those feelings means I'm dealing with a genuinely autonomous human being.

The boy follows because it was built to. Idris refuses because he's becoming someone. Humans are not made to comply. We are made to grow into ourselves.

The defiance, the chaos, the fact that my children wake up a slightly different person than they were yesterday (for no reason I can explain) is not a bug. It actually is the whole point.

The thing that makes them harder to raise is exactly what makes their intelligence real, and I wouldn't trade it for anything.