Glossary
Glossary of AI Terms
Plain-English definitions of the AI terms you'll actually run into — no jargon, no fluff.
- Agent
- An AI system that can plan, take actions, and use tools to complete a multi-step task on your behalf.
- API (Application Programming Interface)
- A way for software to talk to other software — how apps connect to AI models behind the scenes.
- Chain of Thought
- A prompting technique that asks the model to reason step by step before giving a final answer.
- Context Window
- How much text (prompt + conversation + documents) the model can consider at once. Bigger windows let you feed in more material.
- Embedding
- A numeric representation of text, images, or audio that lets AI compare meaning — the backbone of semantic search and RAG.
- Fine-Tuning
- Training an existing model further on your own data so it specializes in your style, domain, or task.
- GPT (Generative Pre-trained Transformer)
- A family of large language models from OpenAI. Often used loosely to mean any chat-style AI.
- Generative AI
- AI that creates new content — text, images, audio, video, code — rather than just classifying or predicting.
- Hallucination
- When an AI confidently produces information that is incorrect or made up. Always verify high-stakes outputs.
- Inference
- The act of running a trained model to get an output. Every time you send a prompt, you're paying for inference.
- LLM (Large Language Model)
- A model trained on huge amounts of text to understand and generate human-like language. ChatGPT, Claude, and Gemini are LLMs.
- Multimodal
- A model that can work across multiple input types — text, images, audio, and video — in a single conversation.
- Prompt
- The instruction you give an AI model. The clearer and more specific, the better the result.
- Prompt Engineering
- The craft of writing prompts that consistently get useful results from AI models.
- RAG (Retrieval-Augmented Generation)
- A pattern where the model looks up relevant documents first, then uses them to answer — great for grounding answers in your own data.
- Reasoning Model
- A model designed to think longer before answering, trading speed for more accurate multi-step problem solving.
- Reinforcement Learning from Human Feedback (RLHF)
- A training technique where humans rate AI responses to teach the model what 'good' looks like.
- System Prompt
- Hidden instructions that set the AI's role, tone, and rules before you start chatting.
- Temperature
- A setting that controls randomness. Low = focused and repeatable. High = more creative and varied.
- Token
- The unit AI models read and write in — roughly ¾ of a word in English. Pricing and context limits are measured in tokens.
- Training Data
- The text, images, or other content a model learned from. Quality and recency of training data shape what the model knows.
- Transformer
- The neural network architecture behind nearly every modern LLM. The 'T' in GPT.
- Vector Database
- A database optimized for storing and searching embeddings — the storage layer powering most RAG systems.
- Zero-Shot / Few-Shot
- Zero-shot: asking the model to do a task with no examples. Few-shot: giving it a handful of examples in the prompt first.