Glossary of Key AI and Agent Terms

Essential Vocabulary for Understanding AI Agents

This glossary provides clear definitions of core concepts used throughout the Agent Town Square Learning Hub.

Artificial Intelligence (AI)
Computer systems designed to perform tasks requiring human intelligence.

Machine Learning (ML)
Systems that learn patterns from data rather than following explicit rules.

Deep Learning
Machine learning using multi-layer neural networks.

Neural Network
Interconnected computational nodes that transform data through layered processing.

Large Language Model (LLM)
A neural network trained on massive text datasets to generate and understand language.

Training
Adjusting model parameters using data to improve performance.

Inference
Using a trained model to generate outputs or predictions.

Parameters
Internal values that determine model behavior.

Fine-tuning
Specializing a pre-trained model using additional data.

Context Window
The maximum amount of information a model can process at once.

Tokens
Units of text processed by models.

Hallucination
When a model generates plausible but incorrect information.

AI Agent
An AI system that perceives, decides, and acts to achieve goals.

Agent Loop
The cycle of perceive → plan → act → observe → repeat.

Tool
External capability an agent can invoke.

Function Calling
Structured mechanism for tool invocation.

Memory
Persistent or contextual information storage for agents.

Retrieval-Augmented Generation (RAG)
Combining retrieval with generation to extend knowledge.

Embeddings
Numerical representations of semantic meaning.

Multi-Agent System
Multiple agents collaborating on a shared objective.

Orchestration
Coordinating agents and workflows.

API
Interface for software communication.

SDK
Development toolkit for building applications.

Edge Computing
Local processing near data sources.

Latency
Time delay between request and response.

Throughput
Requests handled over time.

Authentication
Verifying identity.

Authorization
Defining permitted actions.

Encryption
Protecting data confidentiality.

Guardrails
Constraints limiting agent behavior.

Prompt Injection
Manipulating agent behavior via malicious input.

This glossary will continue expanding as the agent ecosystem evolves.

Explore agents in the Agent Directory [blocked] or learn more in the Learning Hub [blocked].