AI Glossary
Simple explanations of common AI vocabulary for everyone.
A
AI (Artificial Intelligence): Computer systems that can learn, understand, or create things in ways similar to humans.
Agent (AI Agent): An AI system that can take actions, make decisions, and interact with the world or other tools.
Algorithm: A set of rules a computer follows to solve a problem step-by-step.
Alignment: Ensuring AI systems behave safely and follow human values.
API (Application Programming Interface): A way for different software systems to communicate with each other.
Artificial Neural Network: A computer system inspired by the human brain that learns patterns from data.
Automation: Using software or AI to complete tasks without human effort.
Augmented Reality (AR): Digital content overlaid on the real world.
B
Bias (AI Bias): When AI gives unfair or unbalanced results based on flawed data.
Big Data: Extremely large sets of information used to train AI models.
Bots: Automated programs that perform simple tasks online.
Benchmark: A standardized test to evaluate AI performance.
Blockchain: A secure digital record of transactions.
Browser Extension (AI Extension): A small add-on that brings AI features into your browser.
C
Chatbot: A computer program that talks with you using natural language.
Classification: When AI sorts information into categories.
Computer Vision: AI that can understand images and videos.
Context Window: The amount of information AI can remember during a conversation.
Corpus: A large collection of text AI learns from.
Creativity Model: AI systems designed to generate images, music, stories, and art.
Chat Completion: A type of AI output that continues or responds to your message.
D
Data: Information AI learns from, such as text, images, or numbers.
Dataset: A structured collection of data used for training or testing AI.
Deep Learning: A type of machine learning using many layers of neural networks.
Diffusion Model: A type of AI that generates images by turning random noise into pictures.
Domain: A specific area of knowledge or task.
Data Labeling: Tagging data so AI knows what it is.
E
Embedding: A mathematical way AI represents meaning between words or images.
Ethical AI: Building AI that is fair, transparent, and respects human rights.
Explainability: Understanding how an AI made a decision.
Evaluation: Testing how well an AI performs.
Embodied AI
F
Fine-Tuning: Training an AI model on a smaller, specialized dataset.
Foundation Model: A large, general-purpose AI trained on massive data.
Feedback Loop: Using user responses to improve AI over time.
G
Generative AI: AI that creates new content like text, images, or music.
GPU (Graphics Processing Unit): Powerful computer chips used to train or run AI.
Grounding: Ensuring AI answers are based on real facts or data.
H
Hallucination: When AI gives a confident but incorrect answer.
Heuristics: Simple rules AI uses to solve problems quickly.
I
Inference: The process of an AI model generating a response.
Input: The text, image, or instruction you give to an AI.
Image Generation: AI creating pictures from text descriptions.
J
JSON: A simple data format for storing and sharing information.
Judgment Call (AI): When AI makes a subjective choice based on patterns, not facts.
K
Knowledge Base: A library of information an AI can refer to.
Keyword Extraction: AI identifying the main topics in text.
L
Large Language Model (LLM): An AI trained on massive text data to understand and generate language.
Latent Space: A hidden map where AI stores relationships between ideas.
Learning Rate: How quickly AI adjusts during training.
M
Machine Learning: When computers learn from data instead of being manually programmed.
Model: The AI system that learns patterns and generates outputs.
Multimodal: AI that understands multiple types of information (text, audio, images).
Moderation: Filters that prevent harmful or inappropriate AI outputs.
Memory Module
N
Natural Language Processing (NLP): AI that understands and produces human language.
Neural Network: A computer system modeled after the human brain.
Noise (in AI images): The random dots an image AI starts with before creating a picture.
​
O
Output: The result an AI gives you after your prompt.
Optimization: Improving a model’s performance or efficiency.
Overfitting: When an AI memorizes training data instead of learning patterns.
​
P
Parameter: The numbers inside an AI model that control how it makes decisions.
Pattern Recognition: AI detecting similarities in data.
Prompt: The instruction you give to an AI.
​Prompt Engineering: Crafting effective prompts to guide AI.
Q
Query: A question or request sent to AI or a database.
Quantization: Compressing AI models to run faster with less memory.
R
Reinforcement Learning: AI learning by trial and reward.
Retrieval-Augmented Generation (RAG): AI that searches for information before generating answers.
Responsible AI: Building AI that is safe, fair, transparent, and trustworthy.
S
Safety Guardrails: Rules that prevent AI from producing harmful content.
Supervised Learning: Training AI using labeled examples.
Synthetic Data: Artificially generated data used to train AI.
Speech Recognition: AI that turns spoken words into text.
T
Token: Small text pieces AI breaks language into.
Training Data: The examples AI learns from during training.
Transformer Model: A type of AI architecture that excels at language tasks.
Temperature (AI setting): Controls how creative or predictable AI’s responses are.
Tool Use
U
Unsupervised Learning: AI learning patterns without labeled data.
User Intent: What the AI believes you want based on your input.
Upscaling
V
Vision-Language Model (VLM): AI that understands both text and images together.
Voice Cloning: AI copying someone’s voice.
Virtual Assistant: AI that helps with tasks like scheduling or reminders.
W
Workflow Automation: Using AI to complete multi-step tasks automatically.
Weak Supervision: Using imperfect labels to quickly train AI.
Weight (Model Weights): The values a model uses to make decisions.
X
XML: A format for organizing and sharing data.
X-AI (Explainable AI): AI designed to show how it reaches conclusions.
Y
YAML: A simple file format used for AI configuration.
Yield (Model Output Yield): The amount of useful results produced by an AI process.
Z
Zero-Shot Learning: When AI performs a task without being trained on specific examples.
Zooming (in latent space): Exploring deeper layers of AI’s internal “map” of ideas.