AI Vocabulary Cheat Sheet

How to use this sheet

These are the 27 terms you will meet most often in articles, product pages, and conversations about AI. Skim the group you need, or read it top to bottom in a few minutes. For deeper background, start with What Is Artificial Intelligence?

The vocabulary

Core concepts

• Artificial intelligence (AI) — software that performs tasks we associate with human intelligence, such as understanding language, recognizing images, or making decisions.
• Machine learning (ML) — the main way modern AI is built: instead of following hand-written rules, the software learns patterns from examples.
• Model — the trained system itself. When you use a chatbot, a model is doing the work behind the scenes.
• Algorithm — the step-by-step recipe a computer follows. In ML, the algorithm is the learning method; the model is what it produces.
• Neural network — a model loosely inspired by the brain, built from layers of simple connected units that each transform their input a little.
• Deep learning — machine learning that uses neural networks with many layers. It powers today's image, speech, and language systems.
• Large language model (LLM) — a deep learning model trained on huge amounts of text so it can read and write language. ChatGPT-style assistants are built on LLMs.
• Generative AI — AI that creates new content (text, images, audio, code) rather than only classifying or predicting.

How models learn

• Training data — the examples a model learns from. The quality and variety of this data shape everything the model can do.
• Training — the process of adjusting a model, over many passes through the data, until its outputs improve.
• Parameters — the internal numbers a model adjusts during training. "Bigger model" usually means more parameters.
• Supervised learning — learning from labeled examples, such as emails already marked spam or not spam.
• Unsupervised learning — finding structure in unlabeled data, such as grouping customers with similar habits.
• Fine-tuning — taking an already-trained model and training it a bit more on specialized data so it performs better at a narrower task.
• Evaluation — measuring how well a model actually performs, from automated test sets to human review. Every published score is a proxy for real-world usefulness, not a guarantee of it.
• Benchmark — a standardized test set used to compare models. Useful but imperfect: models can effectively memorize public benchmarks, and a high score on exam-style questions may not match your daily tasks.

Using AI tools

• Prompt — the instructions and context you give an AI tool. Better prompts produce noticeably better results.
• Token — a small chunk of text (often part of a word) that a language model reads and writes. Usage and limits are usually counted in tokens.
• Context window — how much text a model can consider at once, including your prompt and its reply. Anything beyond it is effectively forgotten.
• Inference — the moment a trained model is actually used to produce an answer, as opposed to being trained.
• Multimodal — able to work with more than one kind of input or output, such as text plus images or audio.
• API — a way for programs to talk to each other. Apps that "have AI built in" usually call a model through an API.

Risks and limits

• Hallucination — a confident, fluent answer that is simply wrong. The single most important reason to verify AI output.
• Bias — systematic unfairness a model picks up from patterns in its training data, which can skew its answers or decisions.
• Knowledge cutoff — the date after which a model's training data ends. Events past the cutoff are unknown to it unless a tool supplies them.
• Guardrails — rules and filters added around a model to block harmful or off-limits outputs.
• Alignment — the broad effort to make AI systems behave in ways that match human intentions and values.

Where to go next

Once these terms feel familiar, read Large Language Models Explained Simply to see how the pieces fit together, then keep the Machine Learning Terms Cheat Sheet nearby when you hit the more technical vocabulary.