AI Glossary (WIP)

This is a short list of common AI terms and their definitions, compiled as I was becoming familiar with the basics.

Agent

AI system that can take actions autonomously to achieve a goal. Agents can plan and complete complex, multi-step objectives. AI is to consultant as agent is to employee. Agents receive an end goal, figure out how to tackle the job, then complete it. Agents maintain context across actions, rather than losing memory between steps. They work in the observe-think-act agentic loop, observing results and reacting to refine their output.

AI Safety

Field for making AI systems beneficial and safe to use.

Artificial Intelligence (AI)

Development of computer systems that perform tasks typically requiring human intelligence. AI systems learn from data to make decisions, recognize patterns, and solve problems, instead of relying only on explicit, pre-programmed instructions.

Bias

Errors in output resulting from biased training data.

Computer Vision

AI interpreting images and video.

Context

The AI model’s working memory, the information it can access and use when generating a response. Context resets for each conversation, it’s not saved, nor is it training. The context window is the maximum amount of info that a model can hold at once.

General AI (AGI)

Not-yet-achieved AI system with human-level reasoning across any domain.

Hallucination

When a model confidently generates false information.

Inference

Using a model to generate responses in a chat.

Large Language Model (LLM)

AI trained on massive text datasets.

LLMs are a type of ML. ML is a type of AI.

Latency

How long the model takes to respond.

Machine Learning (ML)

Systems that learn patterns from data. Deep learning is a subset of ML using layered neural networks inspired by the brain.

LLMs are a type of ML. ML is a type of AI.

Model

Massive file of numbers (weights or parameters) that work together mathematically to generate responses. The model has no memory, personality, or tools. It generates answers token by token based on probability.

Model Context Protocol (MCP)

Open standard that allows AI models to connect with external tools, data sources, and services. Allows plug-and-play for tools outside conversation with agent. Examples: search the web, access files, query databases, interact with app like Slack or Gmail.

Multi-Agent

Systems where multiple AI models collaborate on a task. Orchestration is coordinating multiple AI agents or steps to complete complex tasks.

Natural Language Processing (NLP)

AI understanding and generating human language.

Prompt Engineering

Crafting inputs to get better AI outputs. Few-shot prompting gives the AI a few examples to learn from in the prompt.

Retrieval Augmented Generation (RAG)

Giving LLM the ability to look up relevant information from an external source before generating a response. Allows to stay up-to-date without retraining.

Skill

Modular set of instructions or best practices that improve AI output for specific a task. A skill is a recipe that the model reads before starting a task. Examples: write text documents, work with PDFs, create UI components, write tests, review pull requests.

Token

Units of text a model processes, used to determine cost to use. One token is about ¾ of a word. Input tokens (text given to model) are lower cost than output tokens (text generated by model).

Training

Process for an AI model to learn from data to develop its capabilities. Billions of times, the model is educated by take in data to make a prediction then using feedback to make better predictions. People fine-tune the model with curated data and use Reinforcement Learning from Human Feedback (RLHF) to produce preferred responses.

*AI was used for research