AI Glossary

Essential Terms for Understanding Artificial Intelligence

This glossary provides clear, accessible definitions of key terms related to artificial intelligence.

Agent / AI Agent

An AI system that can take actions autonomously to accomplish goals. Unlike a simple chatbot that only responds to questions, an agent can browse the web, use tools, make decisions, and complete multi-step tasks on your behalf.

Agentic Workflow

A structured way an AI works through a task by planning steps, taking action, checking results, and adjusting as needed—similar to how a person approaches a project.

AI Literacy

A set of competencies that enables individuals to critically evaluate AI technologies, communicate effectively with them, and understand their ethical and societal implications.

AI Slop

Low-quality, mass-produced content — lazy, cheap, cluttering up the internet. It's often not intentiuonally deceptive, just worthless.

Algorithm

A set of step-by-step instructions that tells a computer how to solve a problem or complete a task. In AI, algorithms are the mathematical recipes that allow systems to learn from data and make predictions.

Algorithmic Bias

Systematic and repeatable errors in a computer system that create unfair outcomes, often privileging one group over others due to skewed training data or flawed objectives.

Alignment / AI Safety

The field focused on ensuring AI systems behave in ways that match human values and intentions. An "aligned" AI does what humans actually want it to do, not just what it was literally told.

Anthropic

An AI safety company founded in 2021 that created Claude. Anthropic focuses on building AI systems that are safe, beneficial, and understandable.

Anthropomorphism

The tendency of users to attribute human-like qualities, consciousness, or emotions to AI systems, often referred to as the "Eliza Effect."

API (Application Programming Interface)

A way for different software programs to communicate with each other. When apps use ChatGPT or Claude behind the scenes, they're connecting through an API.

Artificial General Intelligence (AGI)

A hypothetical AI system that could match or exceed human intelligence across virtually all cognitive tasks. Current AI systems are "narrow"—they excel at specific tasks but can't generalize like humans do.

Artificial Intelligence (AI)

Computer systems designed to perform tasks that typically require human intelligence, such as understanding language, recognizing images, making decisions, and solving problems.

Artificial Narrow Intelligence (ANI)

AI systems designed to solve specific, narrow problems (e.g., chess engines or plagiarism detectors) rather than possessing general reasoning capabilities.

Attention Mechanism

A technique that allows AI models to focus on the most relevant parts of input data. When you ask a question, attention helps the AI determine which words matter most for generating a good response.

Automation

Using technology to perform tasks with minimal human intervention. AI enables more sophisticated automation that can handle complex, variable situations rather than just repetitive tasks.

Bias (in AI)

Systematic errors or unfair outcomes in AI systems, often reflecting biases present in training data or design choices. For example, an AI trained mostly on data from one demographic may perform poorly for others. It is a concern because it can result in discrimination or unfair treatment.

Black Box

An AI system whose internal decision-making process is not transparent or easily understood by humans. You can see the input (what you give the AI) and the output (what it produces), but not the step-by-step reasoning that connects the two. Most deep learning systems, including large language models, function as black boxes—even their creators can't fully explain why they produce specific outputs. This opacity raises concerns about accountability, trust, and the ability to detect errors or bias.

Bot / Chatbot

A computer program designed to simulate conversation with humans. Modern AI chatbots like Claude and ChatGPT use large language models to understand context and generate human-like responses.

Chain-of-Thought (CoT) Prompting

A technique where the user instructs the model to "think step-by-step," forcing it to break down its reasoning process to reduce logic errors

ChatGPT

An AI chatbot created by OpenAI, launched in November 2022. It popularized conversational AI and brought large language models to mainstream public awareness.

Claude

An AI assistant created by Anthropic, designed with a focus on being helpful, harmless, and honest. Claude is known for nuanced understanding and thoughtful responses.

Cognitive Offloading

Using AI to reduce mental effort—such as drafting text, summarizing information, or organizing ideas—so humans can focus on judgment and decision-making.

Confidence vs Accuracy

The tendency of AI to sound very confident even when it is wrong. A confident answer should not be assumed to be correct.

Computer Vision

AI technology that enables computers to interpret and understand visual information from images and videos, including recognizing objects, faces, text, and scenes.

Context Window

The amount of text an AI can consider at once when generating a response. A larger context window means the AI can remember more of your conversation and process longer documents.

Copilot

An AI assistant designed to work alongside humans, particularly in specific applications like coding (GitHub Copilot) or productivity software (Microsoft Copilot).

Data

Information used to train AI systems. This can include text, images, audio, video, numbers, or any other digitized information. The quality and diversity of training data significantly impacts AI performance.

Deep Learning

A subset of machine learning that uses neural networks with many layers to learn complex patterns. Deep learning powers most modern AI breakthroughs, including language models and image recognition.

Deepfake

Deliberately deceptive content — fake videos of real people, manipulated images presented as real, AI-generated "news" designed to mislead. The intent is to deceive, and the content is often sophisticated and convincing.

Embedding

A way of representing words, sentences, or other data as lists of numbers that capture meaning and relationships. Similar items have similar embeddings, allowing AI to understand concepts and find related information.

Emergent Behavior

Unexpected capabilities that appear in AI systems as they become larger or more sophisticated, which weren't explicitly programmed. For example, large language models developed the ability to do math despite not being specifically trained for it.

Ethical AI

The practice of developing and deploying AI systems in ways that are fair, transparent, accountable, and respectful of human rights and values.

Explainable AI

AI systems designed so humans can understand how they reach their conclusions. The opposite of a "black box," explainable AI can show its reasoning, which is important for building trust, catching errors, and meeting regulatory requirements in fields like healthcare and finance.

Few-Shot Learning

An AI's ability to learn a new task from just a few examples provided in the prompt. For instance, showing an AI two or three examples of the format you want before asking it to generate similar content.

Fine-Tuning

The process of further training a pre-trained AI model on specific data to customize it for particular tasks or domains, like medical diagnosis or legal research.

Foundation Model

A large AI model trained on broad data that can be adapted for many different tasks. GPT-4 and Claude are foundation models that can write, analyze, code, and more without task-specific training.

Generative AI (GenAI)

AI systems that can create new content—text, images, music, video, or code—rather than just analyzing or classifying existing content. ChatGPT, Claude, DALL-E, and Midjourney are examples.

Gemini

Google's family of large language models and AI assistants, competing with ChatGPT and Claude. Previously known as Bard.

GPT (Generative Pre-trained Transformer)

A type of AI architecture developed by OpenAI that forms the basis of ChatGPT. The letters stand for: Generative (creates content), Pre-trained (learned from vast data), Transformer (the underlying technology).

GPU (Graphics Processing Unit)

A specialized computer chip originally designed for rendering graphics, now essential for training and running AI models due to its ability to perform many calculations simultaneously.

Grounding

Connecting AI responses to verifiable sources or real-world data. A "grounded" response cites its sources and bases claims on actual information rather than generating plausible-sounding but potentially incorrect content.

Guardrails

Rules and limits built into AI systems to prevent harmful, misleading, or unsafe behavior.

Hallucination

When an AI generates information that sounds plausible but is factually incorrect or completely made up. This is a known limitation of current AI systems and why fact-checking AI outputs is important.

Human-in-the-Loop

A system design where humans review, approve, or correct AI outputs before they're finalized. This provides oversight and catches errors the AI might make.

Image Generation

AI's ability to create pictures from text descriptions. Tools like DALL-E, Midjourney, and Stable Diffusion can generate realistic or artistic images based on written prompts.

Inference

When an AI model uses what it learned during training to process new inputs and generate outputs. Every time you chat with Claude or ChatGPT, the model is performing inference.

Internet of Behavior (IOB)

The Internet of Behavior is an extension of the Internet of Things (IOT) that focuses on understanding and influencing human behavior. It uses data to create personalized and effective ways to improve our daily lives.

Internet of Things (IOT)

The Internet of Things (IOT) refers to a network of physical devices connected to the Internet that can be controlled with our phones or computers. This includes devices like TVs, thermostats, lights, smart home gadgets, etc.

Jailbreaking

Attempting to bypass an AI's safety guidelines or restrictions through clever prompting. AI companies work to prevent jailbreaking to ensure their systems remain safe and helpful.

Knowledge Cutoff

The date after which an AI model has no information, because its training data only extends to that point. AI models cannot know about events that occurred after their cutoff date unless they can search the web.

Large Language Model (LLM)

An AI system trained on massive amounts of text data that can understand and generate human-like language. ChatGPT, Claude, and Gemini are all based on LLMs. "Large" refers to the billions of parameters (learned values) these models contain.

Latency

The delay between when you send a request to an AI and when you receive a response. Lower latency means faster responses.

Machine Learning (ML)

A branch of AI where computers learn patterns from data rather than being explicitly programmed with rules. The computer improves its performance through experience, similar to how humans learn.

Midjourney

A popular AI image generation tool known for producing highly artistic and stylized images from text descriptions.

Model

In AI, a model is the trained system that takes inputs and produces outputs. When you interact with Claude, you're interacting with Claude's model—the result of its training on vast amounts of data.

Multimodal AI

AI systems that can process and generate multiple types of content—such as text, images, audio, and video—rather than just one type. Claude and GPT-4 are multimodal because they can understand both text and images.

Natural Language Processing (NLP)

The field of AI focused on enabling computers to understand, interpret, and generate human language in useful ways. Chatbots, translation services, and voice assistants all use NLP.

Neural Network

A computing system inspired by the human brain, consisting of interconnected nodes (like neurons) that process information. Neural networks are the foundation of modern AI and deep learning.

Open Source AI

AI models whose code and sometimes weights are publicly available for anyone to use, modify, and build upon. Llama (from Meta) and Stable Diffusion are examples of open source AI.

OpenAI

The AI research company that created ChatGPT, GPT-4, and DALL-E. Founded in 2015, it became widely known after launching ChatGPT in 2022.

Output

What an AI system produces in response to your input. This could be text, images, code, audio, or other content depending on the type of AI.

Parameter

A value that an AI model learns during training. Large language models have billions of parameters, which store the patterns and knowledge learned from training data. More parameters generally means more capability.

Perplexity

An AI-powered search engine that provides direct answers with cited sources, rather than just links. It combines search capabilities with conversational AI.

Prompt

The text input you give to an AI system—your question, instruction, or request. The quality of your prompt significantly affects the quality of the AI's response.

Prompt Engineering

The skill of crafting effective prompts to get the best results from AI systems. This includes being specific, providing context, giving examples, and structuring requests clearly.

RAG (Retrieval-Augmented Generation)

A technique where AI retrieves relevant information from a database or documents before generating a response. This helps AI provide more accurate, up-to-date, and sourced answers.

Reasoning

An AI's ability to think through problems logically, break down complex questions, and arrive at conclusions step by step. Advanced models can show their reasoning process.

Reinforcement Learning from Human Feedback (RLHF)

A training technique where AI models learn from human ratings of their responses. Humans evaluate outputs, and the model adjusts to produce responses that humans prefer.

Responsible AI

Developing and using AI in ways that consider societal impact, including fairness, privacy, safety, transparency, and accountability.

Safety (AI Safety)

Research and practices focused on ensuring AI systems don't cause unintended harm. This includes preventing misuse, reducing bias, maintaining human control, and ensuring AI behaves as intended.

Sentiment Analysis

AI's ability to determine the emotional tone of text—whether a message is positive, negative, or neutral. Used in social media monitoring, customer feedback analysis, and more.

Speech Recognition

AI technology that converts spoken words into text. Powers voice assistants like Siri and Alexa, transcription services, and voice-to-text features.

Stable Diffusion

An open-source AI image generation model that can create images from text descriptions. Because it's open source, many apps and tools have been built using it.

Superintelligence

A hypothetical future AI that would greatly surpass human intelligence in virtually all domains. This concept raises important questions about AI safety and control.

Synthetic Data

Artificially generated data used to train AI models when real data is scarce, expensive, or privacy-sensitive. For example, generating fake medical records to train healthcare AI.

System Prompt

Hidden instructions given to an AI that shape its behavior and responses. When you use ChatGPT or Claude, there's often a system prompt working behind the scenes to guide the AI's personality and capabilities.

Temperature

A setting that controls how creative or random an AI's responses are. Lower temperature produces more predictable, focused responses; higher temperature produces more varied, creative ones.

Text-to-Image

AI technology that generates images from written descriptions. You describe what you want to see, and the AI creates a corresponding image.

Text-to-Speech (TTS)

AI technology that converts written text into spoken audio. Modern TTS can produce remarkably natural-sounding voices.

Token

The basic unit of text that AI models process. A token might be a word, part of a word, or punctuation. When AI providers mention "token limits," they're referring to the maximum amount of text the AI can handle.

Training

The process of teaching an AI model by exposing it to large amounts of data. During training, the model learns patterns and relationships that allow it to make predictions or generate content.

Training Data

The information used to teach an AI model. For language models, this typically includes books, websites, articles, and other text. The quality and breadth of training data significantly affects AI performance.

Transformer

The neural network architecture that powers modern large language models. Introduced in 2017, transformers enabled the breakthroughs that led to ChatGPT, Claude, and similar AI systems.

Turing Test

A test proposed by Alan Turing in 1950: if a human can't distinguish between a computer's responses and a human's responses, the computer can be said to exhibit intelligent behavior.

Vector Database

A specialized database designed to store and search embeddings (numerical representations of data). Used to help AI systems quickly find relevant information from large collections of documents.

Vibe Coding

A style of AI-assisted programming where you describe what you want in plain language and let an AI model generate (and iteratively modify) the code, with the human acting more like a director/tester than a line-by-line coder.

Voice Assistant

AI-powered software that responds to voice commands, like Siri, Alexa, and Google Assistant. These combine speech recognition, natural language processing, and text-to-speech.

Weights

The numerical values within a neural network that are adjusted during training. A model's weights encode everything it has learned. When people talk about "model weights," they mean the complete set of learned parameters.

Workflow Automation

Using AI to automatically complete sequences of tasks that previously required human intervention. For example, AI might read emails, extract information, update databases, and send responses.

Zero-Shot Learning

An AI's ability to perform a task without any specific examples—just from general instructions. For instance, asking an AI to translate to a language it wasn't explicitly shown examples for.

Note: The field of AI evolves rapidly. Terms and definitions may shift as technology advances and new capabilities emerge. This glossary reflects common usage as of early 2026.