AI Glossary
Your friendly dictionary of AI terms, explained in plain English. No PhD required.
A
An AI system that can take actions on its own to achieve goals, not just respond to prompts. Agents can use tools, make decisions, and complete multi-step tasks with minimal human guidance.
Example: An AI agent might research a topic, write a report, and email it to you - all from a single instruction.
Technology that enables computers to perform tasks that typically require human intelligence, like understanding language, recognizing images, or making decisions. Think of it as teaching computers to think and learn, though not quite like humans do.
Example: When your phone suggests the next word while texting, that's AI in action.
Application Programming Interface - a way for different software programs to talk to each other. AI APIs let you integrate AI capabilities into your own apps and workflows.
Example: Using OpenAI's API to add ChatGPT features to your website.
C
A prompting technique where you ask the AI to explain its reasoning step-by-step. This often leads to better, more accurate answers for complex problems.
Example: "Solve this math problem and show your work step by step" uses chain-of-thought prompting.
The amount of text an AI can remember and work with at once. Think of it like short-term memory - once you exceed it, the AI starts forgetting earlier parts of your conversation.
Example: If an AI has a 4,000-word context window, it can only consider the most recent 4,000 words of your chat.
F
The process of taking a pre-trained AI model and training it further on specific data to make it better at particular tasks. Like giving a generalist some specialized training.
Example: A company might fine-tune an AI model on their customer service conversations to handle support tickets better.
G
Generative Pre-trained Transformer - the technology behind ChatGPT and similar models. It's a specific type of AI architecture that's particularly good at understanding and generating text.
Example: GPT-4 is the fourth major version of OpenAI's GPT technology.
H
When an AI confidently generates information that's completely made up or incorrect. It's not lying - it just doesn't know the difference between real and fabricated information.
Example: An AI might invent fake statistics or cite non-existent research papers.
L
A type of AI trained on massive amounts of text to understand and generate human-like language. These are the brains behind tools like ChatGPT and Claude.
Example: ChatGPT uses an LLM to have conversations and answer questions.
M
A subset of AI where computers learn patterns from data without being explicitly programmed for every scenario. Instead of following rigid rules, they improve through experience.
Example: Netflix learning what shows you like based on your viewing history is machine learning.
The trained AI system that powers a tool. Different models have different capabilities, sizes, and specialties. Newer models are generally more capable but may cost more to use.
Example: GPT-4, Claude 3, and Gemini are all different AI models.
AI that can understand and work with multiple types of input - like text, images, audio, and video - all at once. More versatile than single-purpose AI.
Example: GPT-4 with vision can analyze images and answer questions about them in text.
N
The branch of AI focused on helping computers understand, interpret, and generate human language. It's what lets AI read your emails, answer questions, and write text.
Example: Autocorrect, voice assistants, and chatbots all use NLP.
A computing system loosely inspired by the human brain, made up of interconnected nodes that process information. The foundation of most modern AI.
Example: Deep neural networks power everything from image recognition to language translation.
P
The text you type to ask an AI tool to do something. A good prompt is clear, specific, and gives the AI enough context to understand what you want.
Example: "Write a professional email declining a meeting invitation" is a prompt.
S
Instructions given to an AI at the start of a conversation that define its behavior, personality, and capabilities. Users typically don't see these, but they shape how the AI responds.
Example: "You are a helpful assistant that speaks like a pirate" could be a system prompt.
T
A setting that controls how creative or random an AI's responses are. Low temperature (0-0.3) makes outputs focused and predictable. High temperature (0.7-1.0) makes them more creative and varied.
Example: Use low temperature for factual answers, high temperature for creative writing.
The basic unit of text that AI models process. Roughly, one token equals about 4 characters or 0.75 words. AI pricing and limits are often measured in tokens.
Example: The sentence "Hello, world!" is about 4 tokens.
The massive collection of text, images, or other information used to teach an AI model. The quality and diversity of training data heavily influences what the AI can do.
Example: An AI trained on medical textbooks will be better at health questions than one trained only on novels.
Z
Zero-shot means an AI can perform a task without any examples. Few-shot means it needs just a handful of examples to understand what you want. Both show how flexible modern AI is.
Example: Asking an AI to translate to a language it wasn't specifically trained for is zero-shot learning.