AI FAQ & Glossary

A

Algorithm

A step-by-step set of instructions for solving a problem. In AI, algorithms are the "recipes" that learn patterns from data.

Artificial Intelligence (AI)

The broad field of building machines that perform tasks we associate with human intelligence, such as understanding language or recognizing images.

Artificial General Intelligence (AGI)

A hypothetical future AI that could match human ability across virtually any task. Today's AI is "narrow" — good at specific things, not everything.

API

Application Programming Interface — a doorway that lets one program talk to another. Developers use AI companies' APIs to add AI features to their own apps.

B

Bias

Unfair or skewed behavior that a model learns from imbalanced or prejudiced training data. Reducing bias is a core part of responsible AI.

Big Data

Datasets so large or complex that ordinary tools struggle to handle them. Big data is much of the fuel behind modern AI.

C

Chatbot

A program you interact with through conversation. Modern chatbots are powered by large language models.

Classification

A task where a model sorts inputs into categories — e.g. "spam" vs. "not spam," or which animal is in a photo.

Clustering

Grouping similar items together without being told the categories in advance. A common form of unsupervised learning.

Computer Vision

The field of teaching computers to interpret images and video — recognizing faces, reading handwriting, spotting tumors in scans.

D

Data

The examples an AI learns from — text, images, numbers, sounds. Quality and variety of data strongly shape how well a model performs.

Dataset

An organized collection of data used to train or test a model.

Deep Learning

A powerful branch of machine learning that uses neural networks with many layers, enabling breakthroughs in vision and language.

Diffusion Model

A type of model behind many image generators. It learns to turn random noise into a coherent picture, step by step.

E

Embedding

A way of turning words, images, or other items into lists of numbers so a computer can measure how similar they are. Similar things end up "close together."

Epoch

One full pass through the entire training dataset during learning. Models usually train for many epochs.

F

Feature

An individual measurable property the model pays attention to — e.g. the price, size, or color of something.

Fine-tuning

Taking an already-trained model and giving it extra training on a narrower topic so it specializes for a specific job.

G

Generative AI

AI that creates new content — text, images, audio, code — rather than just classifying or predicting existing things.

GPT

Generative Pre-trained Transformer — a family of large language models. The technology behind tools like ChatGPT.

GPU

Graphics Processing Unit — a chip originally built for video games, now essential for training AI because it handles huge numbers of calculations in parallel.

H

Hallucination

When an AI states something false but plausible-sounding as if it were fact. Always verify important AI answers.

Hyperparameter

A setting chosen before training that shapes how a model learns (like an oven's temperature dial before you bake).

I

Inference

Using a trained model to make a prediction or generate an answer. Training is the studying; inference is answering the question.

Image Recognition

The ability of a model to identify what's in a picture — a subtask of computer vision.

L

Label

The "correct answer" attached to a training example — e.g. tagging a photo as "cat." Labeled data powers supervised learning.

Large Language Model (LLM)

A model trained on vast amounts of text to understand and generate human-like language by predicting likely next words.

M

Machine Learning (ML)

The approach of getting computers to learn patterns from data instead of being explicitly programmed with rules.

Model

The trained system that results from machine learning — the "expert" that makes predictions or generates content.

Multimodal

An AI that works across more than one kind of input or output at once — for example, understanding both text and images.

N

Natural Language Processing (NLP)

The field of helping computers understand and produce human language — behind translation, chatbots, and voice assistants.

Neural Network

A model loosely inspired by the brain, built from layers of connected "neurons" that pass signals along to detect patterns.

Neuron (Node)

A single tiny unit in a neural network that receives inputs, does a small calculation, and passes the result on.

O

Overfitting

When a model memorizes its training data too closely and then performs poorly on new data — like a student who memorized answers instead of learning the subject.

Open Source

Software (or a model) whose code or weights are publicly available for anyone to inspect, use, and build on.

P

Parameter

An internal value a model adjusts during training. Large models have billions of them — together they store what the model has "learned."

Pattern Recognition

Spotting regularities in data — the core thing most AI actually does.

Prompt

The instruction or question you give an AI. Clear, specific prompts usually get better results.

Prompt Engineering

The skill of wording prompts well to get the output you want from an AI tool.

R

Reinforcement Learning

Training by trial and error, where a model earns rewards for good actions — how AI learns to play games or steer robots.

S

Supervised Learning

Learning from labeled examples where the correct answers are provided — the most common form of machine learning.

T

Token

A small chunk of text (a word or piece of a word) that language models read and generate. AI usage is often measured in tokens.

Training

The process of showing a model many examples so it can adjust its parameters and learn patterns.

Transformer

The neural-network design that powers most modern language models. Its "attention" mechanism lets it weigh which words matter most.

Turing Test

A classic thought experiment: can a machine hold a conversation well enough that you can't tell it apart from a human?

U

Unsupervised Learning

Finding structure in data without labeled answers — e.g. grouping customers by behavior. Clustering is a common example.

Underfitting

When a model is too simple to capture the real pattern, so it performs poorly even on its training data.

W

Weights

The adjustable strengths of the connections in a neural network. Training is essentially the process of tuning the weights.