AI Terminology Glossary

This glossary explains the AI terms you are most likely to see in product pages, research summaries, and business conversations.

AGI

Artificial general intelligence: a hypothetical AI system that can perform broadly across domains at human-level flexibility. No current public AI system is generally accepted as AGI.

AI Agent

An AI system that can use tools, follow steps, and pursue a goal beyond one simple response.

Alignment

The work of making AI systems behave according to human intentions, safety expectations, and stated constraints.

Attention

A mechanism that helps a model weigh which parts of the input are most relevant.

Benchmark

A standardized test used to compare AI models. Benchmarks are useful, but they do not always predict real-world performance.

Context Window

The amount of text, code, or other input a model can consider at once.

Embedding

A numerical representation of text, images, or other content. Embeddings make similarity search possible.

Fine-Tuning

Additional training that adapts a model to a specific task, style, or data pattern.

Generative AI

AI that creates content such as text, images, audio, video, or code.

Hallucination

When an AI produces false or unsupported information while sounding confident.

Inference

Using a trained model to generate an output.

LLM

Large language model. A model trained on large text or code datasets to understand and generate language.

Machine Learning

An AI approach where systems learn patterns from data instead of only following hand-written rules.

Multimodal AI

AI that can process more than one type of input, such as text, images, audio, video, or code.

Prompt

The instruction or input you give to an AI model.

RAG

Retrieval-augmented generation. A pattern where a system retrieves relevant source material before asking a model to answer.

RLHF

Reinforcement learning from human feedback. A training method that uses human preference signals to improve model behavior.

Token

A piece of text processed by a model. Tokens can be words, word parts, punctuation, or characters depending on the tokenizer.

Transformer

The neural network architecture behind most modern language models. Transformers use attention to process context efficiently.

Vector Database

A database designed to store embeddings and search by similarity. Commonly used in RAG systems.

Zero-Shot

When a model performs a task without being given examples in the prompt.

Verified Sources