Thanks Yuki, I didn't know Randi Williams.
Absolutely love the example of this child training an AI to play rock-paper-scissor, then mischievously training it the wrong rules so that they'd win, thus understanding that... humans are actually the ones training AIs.
Inviting our children to question the answers we've being given, through - again -conversation. "Why is this the answer we are getting?" "Do we agree or disagree?" "Who do we think wrote the rules?"
As she puts it in the end : "Children need a model to know what it means to be a curious user of AI."