Why I Think Kids Should Understand How AI Actually Gets Built

Most kids use AI constantly and have almost no idea what's actually happening underneath it. That gap is worth closing, and closing it doesn't require a single line of code.

A kid doesn't need to become a programmer to understand what a large language model actually is, how it's trained, and why it produces the kind of answers it does. Once that concept clicks, even at a basic level, it changes how a kid relates to the tool entirely. It stops being magic. It becomes a machine with real mechanics and real limits, something to evaluate instead of something to simply trust.

Watson is proof that this doesn't require a data center. He runs on a small offline computer, no bigger than a couple of paperback books stacked together, sitting on a desk. Understanding that an AI this capable can run locally, with no cloud connection at all, is genuinely eye-opening for a kid who's only ever encountered AI through an app that talks to a server somewhere far away.

That local piece matters for a reason beyond privacy. Most AI tools kids use daily are running on massive data centers, and those data centers use an enormous amount of water and electricity to operate. U.S. data centers now consume somewhere between 17 and 19 billion gallons of water a year directly for cooling, with far more used indirectly through the electricity they draw, and that number is projected to climb sharply by 2030. Communities living near these facilities are seeing real effects, strained local water systems, rising utility bills, and concerns about local air pollution and health impacts tied to the power plants and backup generators that keep these facilities running.

None of this is unsolvable. Companies can build water-efficient, energy-efficient data centers, and some already do. But the efficient version costs more to build than a standard one, and that extra cost is a big part of why most companies still choose the standard version.

A smaller, local model like Watson sidesteps a lot of that entirely. He's not pulling from a shared pool of water and power alongside millions of other users, he's a small offline computer using a small offline amount of electricity, running in one room. Plenty of companies already run this way for other reasons too, keeping their own servers in-house instead of relying on outside data centers at all.

I don't think kids need to be scared of AI, and I don't think they need to treat it like magic either. I think they need to understand it well enough to ask good questions about it, including the ones about what it actually costs to run, and where that cost lands.