The AI glossary
Plain-English definitions of the AI terms you'll meet while choosing tools — no jargon, no hype. 53 terms across 6 topics.
Fundamentals
Software that performs tasks normally needing human intelligence — understanding language, recognizing images, making decisions.
A way of building software that learns patterns from data instead of being explicitly programmed.
Machine learning that uses many-layered neural networks to learn complex patterns.
A model loosely inspired by the brain, made of connected layers of simple units that transform data.
The process of teaching a model by showing it data and adjusting it to reduce errors.
Running a trained model to get an output — the part that happens when you use an AI tool.
The internal numbers a model learns during training; more can mean more capacity.
Language models
A model trained on huge amounts of text that can generate and understand language.
The neural-network architecture behind most modern language and multimodal models.
A chunk of text (roughly a word-piece) that language models read and generate.
The maximum amount of text a model can consider at once, measured in tokens.
A setting that controls how random or creative a model's output is.
The input or instruction you give an AI model to get a response.
The practice of writing and refining prompts to get better results from AI models.
Prompting a model to reason step by step, which improves accuracy on hard problems.
Asking a model to do a task with no examples — just the instruction.
Giving a model a handful of examples in the prompt to guide its output.
A model trained or prompted to "think" longer before answering, for harder problems.
The date after which a model has no built-in knowledge of world events.
Generative AI
AI that creates new content — text, images, audio, video, or code.
The technique behind most AI image and video generators.
An older image-generation method where two networks compete to produce realistic output.
Generating images from a written description.
Generating short video clips from a prompt or image.
Turning written text into natural-sounding spoken audio.
Transcribing spoken audio into written text.
Recreating a specific person's voice from a short audio sample.
A model that works across more than one type of input — text, images, audio, video.
Synthetic media that realistically depicts a real person saying or doing something they didn't.
Building with AI
Giving a model relevant documents at query time so answers are grounded in your data.
A numeric representation of text or images that captures meaning for search and matching.
A database built to store embeddings and find the most similar items fast.
Further training a base model on your own data to specialize it.
A large, general model trained on broad data that others build on top of.
A model whose weights are publicly available to run, modify, and self-host.
An interface that lets one program use another — how apps call AI models.
Letting a model trigger real actions or fetch data by calling defined tools.
An AI system that plans and takes multi-step actions toward a goal, not just one reply.
AI that acts with autonomy — making decisions and taking actions across steps.
An open standard for connecting AI models to external tools and data sources.
Training method that uses human preferences to make models more helpful and aligned.
Training where a model learns by trial and error from rewards.
Performance & infrastructure
Shrinking a model by using lower-precision numbers so it runs faster and cheaper.
The delay between sending a request and getting a response.
AI that interprets images and video — detecting objects, reading text, understanding scenes.
The field of AI focused on understanding and generating human language.
Artificially generated data used to train or test models when real data is scarce.
Safety & ethics
When an AI confidently states something false or made-up.
Making sure an AI system's behavior matches human intentions and values.
Rules and filters that keep AI output safe, on-topic, and within policy.
Systematic unfairness in AI output that reflects skews in its training data.
When a model memorizes its training data and fails to generalize to new inputs.
Embedding a hidden signal in AI output to mark it as AI-generated.
Now find the tools
Put the terms to work — browse hand-reviewed AI tools across every category.