AI Glossary

Original, plain-English definitions of the terms you'll run into when reading or talking about AI.

AI Agent: An AI system that can take multiple steps toward a goal on its own — deciding what to do next, using tools, and adjusting based on results — rather than just producing a single response to a single prompt.; Related:Large Language Model (LLM)Inference
Artificial Intelligence (AI): Software designed to perform tasks that normally require human intelligence — understanding language, recognizing images, or making decisions. It's a broad goal, not one specific technique.; Related:Machine Learning (ML)Generative AI
Context Window: The maximum number of tokens a model can consider at once, including your input, any provided files, and its own reply. Content beyond this limit gets dropped or summarized.; Related:Token Large Language Model (LLM)
Deep Learning: A type of machine learning that uses layered mathematical structures called neural networks, loosely inspired by how neurons connect in the brain. It's the technique behind most modern language and image models.; Related:Machine Learning (ML)Large Language Model (LLM)
Embeddings: A way of representing the meaning of text (or images) as a list of numbers, positioned so that similar meanings end up close together. This lets computers measure how related two pieces of content are.; Related:Retrieval-Augmented Generation (RAG)
Few-Shot Learning: Giving a model a small number of examples directly in your prompt before asking it to do the task, so it can pick up the pattern you want (format, tone, style) without any retraining.; Related:Zero-Shot Learning Prompt Engineering
Fine-Tuning: Further training an already-trained model on a smaller, specific dataset so it performs better on a narrower task or adopts a particular style, without starting from scratch.; Related:Inference Machine Learning (ML)
Generative AI: AI systems built to produce new content — text, images, audio, or code — rather than just classify or predict a number. Chatbots and image generators are both generative AI.; Related:Large Language Model (LLM)Artificial Intelligence (AI)
Hallucination: When an AI model states something false, invented, or unsupported with the same fluent confidence as a correct answer. It happens because models generate statistically plausible text, not verified facts — always worth double-checking anything factual and important.; Related:Large Language Model (LLM)Inference
Inference: The process of a trained AI model actually producing an output for a given input — as opposed to "training," which is when the model is first built from data. Every time you send a prompt and get a reply, that's inference.; Related:Fine-Tuning Large Language Model (LLM)
Large Language Model (LLM): A deep learning model trained on huge amounts of text to predict what text should come next. This simple ability, at large scale, lets it answer questions, write, summarize, and hold conversations.; Related:Deep Learning Generative AI Token
Machine Learning (ML): A way of building AI systems by showing them many examples and letting them find statistical patterns, rather than hand-coding every rule. Most modern AI is built this way.; Related:Artificial Intelligence (AI)Deep Learning
Multimodal AI: An AI model that can understand or generate more than one type of content — for example, reading both text and images in the same conversation, rather than being limited to text alone.; Related:Generative AI Large Language Model (LLM)
Neural Network: A mathematical structure made of layered, connected nodes that loosely mimics how neurons connect in a brain. Data passes through the layers, and each connection's strength is adjusted during training until the network produces useful output.; Related:Deep Learning Parameter
Parameter: One of the internal numbers a model adjusts during training to get better at its task. Modern LLMs have billions of parameters — when you see a model described as "70 billion parameters," that's the count of these adjustable numbers, roughly indicating its scale.; Related:Neural Network Machine Learning (ML)
Prompt Engineering: The practice of deliberately structuring the instructions you give an AI model — including role, task, context, and format — to reliably get higher-quality, more predictable output.; Related:Large Language Model (LLM)Context Window
Retrieval-Augmented Generation (RAG): A technique where a system first searches a set of documents for relevant information, then hands that information to a language model as extra context before it answers — improving accuracy on specific or private information.; Related:Embeddings Large Language Model (LLM)
System Prompt: Background instructions given to an AI model before your conversation starts, usually set by the application rather than you — defining its role, tone, or rules (e.g., "you are a helpful customer support agent"). You don't see it, but it shapes every reply.; Related:Prompt Engineering Large Language Model (LLM)
Temperature: A setting that controls how predictable or varied a model's output is. Low temperature makes it favor the single most likely next word every time (more focused, repeatable); high temperature lets it pick less-likely words more often (more varied, sometimes less coherent).; Related:Inference Large Language Model (LLM)
Token: The small chunk of text — a word, part of a word, or punctuation mark — that a language model reads and generates one unit at a time. In English, one token is roughly three-quarters of a word.; Related:Context Window Large Language Model (LLM)
Training Data: The large collection of examples — text, images, or other content — a model learns patterns from during training. A model's abilities and blind spots both trace back to what was and wasn't in its training data.; Related:Machine Learning (ML)Fine-Tuning
Transformer: The neural network design introduced in 2017 that made modern LLMs possible. Its key trick — called "attention" — lets the model weigh how relevant every other word in the input is to each word it's processing, which handles long-range context far better than earlier designs.; Related:Large Language Model (LLM)Neural Network
Zero-Shot Learning: Asking a model to perform a task it was never specifically shown examples of during training or in your prompt — relying entirely on what it learned generally. Most everyday chatbot use is zero-shot.; Related:Few-Shot Learning Large Language Model (LLM)

New to these terms? Start with the AI Fundamentals course →