A short glossary for confused parents

Your kid comes home saying "the model kept hallucinating" or "I had to fix my prompt" and you nod along, because asking her to explain feels like admitting you're behind. You're not behind — most of these terms are recent, and nobody handed out a glossary. Here's one, short and plain, no assumed background required.

Model

A model is the actual "brain" — a program trained on huge amounts of text, images, or other data to notice patterns, then used to generate answers, predictions, or content. When your kid says "the model," she means the specific AI system doing the work, the way you'd say "the calculator" instead of "math."

LLM

Large Language Model — the specific kind of model behind most chatbots. It's trained on enormous amounts of text and learns to predict, one piece at a time, what words are likely to come next given what came before. That's a simplification, but it's the useful one: an LLM is a very sophisticated next-word guesser, not a mind that knows things the way a person does.

Prompt

The instruction or question you give a model. "Write me a story about a robot dog" is a prompt. A big part of what kids practice here is prompting well — being specific about what you want, instead of vague — because a vague prompt reliably gets a vague, less useful answer.

Hallucination

When a model states something confidently that isn't true. It's not lying in the human sense — it doesn't know it's wrong — it's a known limitation of how these systems generate answers. We teach kids to expect this and check outputs rather than trust them blindly; a hallucination caught by a kid is a genuinely good learning moment, not a failure of the session.

Sensor

Any component that lets a machine notice something about the physical world — light, distance, sound, temperature, motion. We've written a full piece on sensors as "how machines feel" if you want the longer version; the short version is that a sensor is the machine's only way of knowing anything true about what's actually happening around it.

Training

The process of teaching a model by showing it enormous amounts of example data until it gets good at recognizing patterns in that data. A kid doesn't usually train a model from scratch here — that takes resources way beyond a classroom — but she might fine-tune or steer one, or train a small, simple model of her own on a narrow task like image recognition.

Inference

What happens when a trained model actually gets used — you give it an input, it processes that input using what it learned during training, and produces an output. If training is studying for the test, inference is taking the test.

Token

The small chunks — often pieces of words — that a language model actually reads and generates, one at a time. Not a term your kid needs to use fluently, but it's why longer prompts and longer answers cost more and take longer: more tokens, more work.

Why we're not defining "AI" itself

It's the one term in this list broad enough to mean almost anything depending on who's using it, and pinning it down precisely would take longer than the rest of this glossary combined. If your kid uses it loosely, that's normal — everyone does, including us. The specific terms above are the ones actually worth pinning down, because they're the ones that show up in the real, concrete decisions she's making while she builds.