Key AI Terminology Explained

AI has its own vocabulary. Here are the terms you will encounter most often, explained in plain language.

Model

A model is the trained AI system itself. It takes an input and produces an output. When people say "GPT-4" or "Claude," they are referring to specific models.

Training

Training is the process of teaching a model by showing it data. A language model is trained on text; an image model is trained on images. Training happens once (or periodically) and requires significant compute resources.

Inference

Inference is when a trained model processes a new input and generates a response. Every time you ask ChatGPT a question, that is inference. Inference costs money per use, which is why AI pricing is often per-token or per-request.

Tokens

Tokens are the units that language models use to process text. A token is roughly three-quarters of a word. The sentence "The quick brown fox" is about five tokens. Token counts matter because they determine cost and the amount of text a model can process at once.

Context Window

The context window is the maximum amount of text a model can consider at one time. A model with a 128K token context window can process roughly 96,000 words in a single request. Larger context windows let you include more information but cost more.

Hallucination

A hallucination is when an AI generates information that sounds plausible but is factually incorrect. Models do not "know" facts — they predict likely text. This means they can confidently state things that are wrong.

Fine-Tuning

Fine-tuning is customizing a pre-trained model with your own data to make it better at a specific task. It is more involved than prompting but less expensive than training from scratch.

Embedding

An embedding is a numerical representation of text (or images, or other data) that captures its meaning. Embeddings let you compare how similar two pieces of content are, which powers search and recommendation systems.

RAG (Retrieval-Augmented Generation)

RAG is a pattern where you search your own data first, then include the relevant results in a prompt to the AI model. This lets the model answer questions using your specific information rather than just its training data.

API (Application Programming Interface)

An API is how software systems talk to each other. When a business integrates AI, it typically calls an AI provider's API — sending a request and receiving a response programmatically.