AI Glossary — plain-English terms for parents

Every NextGen Kids ritual uses these words. Here's what they actually mean — in language you can repeat to your kid.

AI (Artificial Intelligence)

A computer program that's been trained to predict what should come next — a word, a pixel, a sound. That's it. There's no thinking, no understanding, no awareness. Just very, very fast prediction. When AI "answers" your question, it's predicting what a helpful answer would look like, based on millions of examples it's read.

Chatbot

An AI you can talk to in regular sentences. ChatGPT, Gemini, Claude, and Copilot are all chatbots. They reply in real time, in conversational language. They're not looking up answers like Google — they're generating them, word by word.

Model

The actual AI program — the trained thing that does the predicting. Companies release "models" the way car companies release car models. GPT-4 is a model. Gemini Flash is a model. The model is what's underneath the chatbot interface.

Prompt

What you type into a chatbot. The clearer your prompt, the better the answer. "Help with maths" is a bad prompt. "My 8-year-old is stuck on this question — explain step by step in simple words" is a good prompt.

Hallucination

When AI confidently says something untrue. This is not a bug — it's how AI works. It predicts what an answer should sound like, even when it doesn't actually know. Spotting hallucinations is one of the most important skills your kid will build.

Training data

The text, images, and code AI was shown during its training. Most modern AI has been trained on a huge slice of the internet. That's why it can answer in many languages, knows about many topics, and also why it can be wrong, biased, or outdated.

Tokens

The unit AI works in. A token is roughly a word — or a piece of one. When you hear "this AI handles 100,000 tokens of context," that means it can read about 75,000 words in a single conversation before forgetting.

Bias

The systematic ways AI's training data leans in one direction. AI trained mostly on English-language internet sources will be better at English-speaking contexts and worse at Indian ones, for example. Bias isn't malicious — it's structural.

Open-source

An AI whose model is published publicly so anyone can run it on their own computer. ChatGPT is closed-source. LLaMA from Meta is open-source. The distinction matters for trust and privacy.

API

The plumbing developers use to plug AI into other apps. When your bank's chatbot or your school's homework helper "uses AI," they're usually calling an API behind the scenes. You won't typically interact with APIs directly — but it's useful to know the word exists.

Multimodal

AI that handles more than just text. ChatGPT can now read images, generate them, and listen to your voice. That's multimodal. Most modern AI is moving in this direction.

Agent

An AI that doesn't just answer questions — it takes actions. Booking a flight, sending an email, browsing a website. Agents are still early and often unreliable. They're worth knowing about because they're where the technology is headed in 2026.

More terms? Write to us and we'll add them.