Pamela Fox published her slide deck on what we need to change in the way we teach CS all across the different age ranges, starting from primary school and going up to the university degree.

Some of my colleagues were in disbelief that she included "Understanding how AI works" as a primary school objective.

Just like you can't prevent teenagers from lying about their birthdate and getting access to porn, you will not be able to prevent kids from using AI models, especially if Big Tech has their way with it and every product in your household is infused with a chatbot/agent.

The best way to make AI safe for kids is through education. You need to teach kids how AI works, and what its limits are and what you can do safely or not.

Dale Lane has this awesome project called Machine Learning for Kids where kids can write ML and AI applications using a scratch-like programming environment.

In today's video, Dale goes through jailbreaking, where the agent is supposed to keep a password protected, and the user tries different things to extract that password from a weak-willed agent.

So yes, you can teach primary school kids how AI works. And we should be doing this right now, not in a couple of years.