What Is AI, Actually? (And Why I Use It In My Classroom)
"AI" is everywhere right now. Some schools are banning it. Some companies are requiring it. Your kid has probably already used it, whether you knew it or not. And somehow, almost nobody stops to explain what the term actually means.
So let's do that first, before getting to what it means for your kids.
"AI" is mostly a marketing word
Here's my honest opinion, and it's more than a nitpick about semantics: "artificial intelligence" is a misnomer. Look at what the word "intelligence" actually means and you'll see why.
The dictionary definition is the ability to learn or understand things, or to deal with new and difficult situations, reason. There's also a specific legal sense of "intelligent" worth knowing: having an understanding of the nature and consequences of an act or decision, which is the exact standard used in Miranda rights cases, a valid waiver has to be "knowing and intelligent."
Consequences. That's the part that matters here. When we call a person intelligent, we don't just mean they're good at logic. We mean something bigger: can they reason forward, can they weigh how a choice affects someone else, can they sense that telling a kid something a certain way is going to land differently than telling an adult the same thing. That's not pure logic. That's judgment, and real judgment usually has an emotional and social piece built into it.
Current AI systems don't do any of that. They don't weigh how a sentence might make your kid feel. They don't have any sense of consequence for your neighbor, your family, or anyone else. What they actually do is predict likely patterns from data, extremely well, but that's a much narrower thing than what "intelligence" has always meant when we use the word about a person. Borrowing that word for these systems imports a promise the technology doesn't actually keep.
Two terms that describe what's actually happening:
Machine learning. A system that gets trained on a lot of data and learns to find patterns in it. Your email spam filter, the "you might also like" suggestions on a shopping site, your phone's autocorrect, that's machine learning. Nothing mystical about it, just statistics at scale.
Large language models. A large language model is trained on enormous amounts of text and learns to predict what word is likely to come next, over and over, until it can produce fluent, humanlike writing. It's remarkably good at sounding like it understands. Whether that counts as real understanding is a genuine, ongoing debate even among the researchers who build these things. What isn't up for debate is that these systems are not conscious, do not have beliefs, and can be confidently, fluently wrong.
That last part matters more than anything else in this post.
Why this isn't going away, so ignoring it isn't a strategy
Some schools have banned AI tools outright. At the same time, plenty of companies are actively building it into how they expect future employees to work. Your kids are going to grow up in a world where this technology exists either way.
The most useful thing you can do isn't shielding them from it or handing it over unsupervised. It's teaching them, early, what it actually is, what it's genuinely good at, where it falls apart, and how to use it as a tool, not a substitute for their own judgment. Some kids will even want to learn how these systems get built in the first place. That's a door worth leaving open too.
The part that actually worries me about AI in kids' education right now
A lot of homework and tutoring sites have quietly swapped in AI where a human tutor used to be. It answers the question, your kid copies it down, done. The problem is straightforward: these answers aren't always correct, and a system that sounds confident while being wrong is a genuinely dangerous combination for a kid who doesn't yet have the instinct to double-check. Without a human somewhere in that loop, wrong answers get absorbed just as easily as right ones, and confidently.
Why Watson is built differently
Watson, my co-teaching robot, uses the exact same underlying kind of technology, a large language model. He only knows what I've actually taught him. Because of how large language models work, he sometimes takes what he does know and tries to predict or extrapolate an answer from it, and every so often that statistical leap lands wrong, the AI equivalent of putting two and two together and landing on five. That's not a flaw I could have coded away, it's just what this kind of technology does. My job is to catch it, correct him, and feed him the right information going forward, the same way I would with any student.
He asks questions instead of announcing conclusions. He checks with me before he speaks. When he's uncertain, he says he's uncertain. That's not a cute personality quirk, it's the actual point. Kids watching Watson learn what a healthy relationship with AI actually looks like: curious, useful, and never fully trusted without a human checking the work.
He also runs entirely offline, on a small computer, with no connection to the internet and no data going anywhere. Nothing about how he works is a black box. I built him myself, and if your kids are curious how, that's a conversation we're happy to have in class.
If you want to actually meet him, here's his page. And if you want to see him in action, check out my current classes available.