Key AI terms, clearly explained
A practical reference for the concepts behind modern AI tools — from the foundational (LLMs, tokens) to the technical (RAG, fine-tuning) to the emerging (MCP, agents).
Foundation
How AI models focus on the most relevant context
AI that can see and interpret images and video
AI learning with many layers of neural networks
The technology behind AI image generation
Converting meaning into numbers AI can compare
The large, general-purpose model everything builds on
The engine behind most modern AI tools
A model architecture that activates only part of its capacity per token
AI that understands text, images, and more at once
Teaching computers to understand human language
The computing architecture inspired by the human brain
The breakthrough design that powers modern AI
Concepts
AI that takes actions, not just answers questions
Making AI goals match human values
Programmatic access to AI models for developers
How much an AI can "remember" in one conversation
AI that creates new content — text, images, audio, video
Connecting AI outputs to verifiable facts
Rules and filters that prevent AI from harmful outputs
When AI confidently states something false
Running a trained AI model to generate output
The date after which an AI model has no information
A standard for connecting AI to tools and data
The numbers that define what a trained AI model knows
Multiple AI agents collaborating to solve complex tasks
AI models whose weights are publicly available
Tricking AI by hiding malicious instructions in content
Displaying AI output token-by-token as it is generated
The hidden instructions that shape AI behavior
The dial that controls AI creativity vs. consistency
The basic unit of text that AI models process
The text and examples that shape an AI model
Techniques
Getting AI to think step-by-step
Training AI to be helpful, harmless, and honest via rules
Adapting a model to a specific task or domain
AI adapting to tasks from examples in the prompt
Training a small model to mimic a large one
Shared vector space for text, images, and other data types
The art of talking to AI models effectively
Making large AI models smaller and faster
AI learning by trial-and-error with reward signals
Giving AI models access to your own data
How AI learns to be helpful and safe
Search that understands meaning, not just keywords
AI-generated training data used to train other AI
Letting AI call external APIs and services
Storage built for AI-powered semantic search
Teaching AI by example — or without any examples