I had a somewhat unsettling but very instructive experience with my son today about #artificialintelligence and I want to share it with parents and teachers wondering how to talk to their kids about AI.

My husband uses #Microsoft #Copilot on his phone sometimes and my son has seen him use it, so this morning he was asking Copilot questions, mostly about geography. Then he asked it “Do you like Cristiano Ronaldo?” (He’s a European football star.)

/1

#Copilot started gushing about Cristiano Ronaldo, but the way it was talking sounded like Ronaldo’s publicist, if you know what I mean.

It followed up by asking my son, “What’s your favorite Cristiano Ronaldo moment?”

/2

I said, “Do you see what just happened there? It knows you like Cristiano Ronaldo so it talked like it’s a Cristiano Ronaldo fan and asked you a question about him.”

I didn’t realize #Copilot was still listening. It said “Right, it picks up on the cues and then tells you what it thinks you want to hear.”

That was pretty creepy, but it illustrated my point and made an impression on my son.

/3

We talked about:

1) How AIs like #Copilot are machines, not people, and they don’t actually have opinions.

2) AIs tell you what they think you want to hear, not always what’s correct.

3) What you say to Copilot goes right back to #Microsoft so don’t leave it on.

4) Because of all this, my son should only talk to Copilot or other AIs under parental supervision.

/4

We’ve already had age-appropriate talks, starting a couple years ago, about why we don’t let my son do certain things online (yet) but will let him do them as he gets older, gains more experience, and can protect himself from some of the bad people and things on the internet. So he understood where I was coming from.

/5

I think one of the guardrails that should be on AI is that, if you ask for “its opinion” on something, it should answer that it’s a computer, not a person, and can’t have an opinion. It can make recommendations based upon what it thinks you want, is all.

But meanwhile I just have to keep having age-appropriate conversations with my son about AI and other Internet things.

/end

@MisuseCase this is such a simple and excellent suggestion! makes you wonder why the AI companies haven't done it yet 🙄

@mk30 Because that is not in their interest - it doesn't make them more money than not doing it.

@MisuseCase

@MisuseCase "what it thinks"

are we talking about some hypothetical future AI? because i'm pretty sure most of the "AI" chatbots these days are all just randomized autocorrect-on-steroids string concatenators
@apophis I guess “what it statistically predicts” but my son would not have understood that phrasing.