Agent
An LLM that plans, calls tools, and acts toward a goal.
Example: An agent books a meeting and emails the team.
AI Dictionary
AI terms in plain language — search and learn.
Algorithm
A step-by-step procedure a computer follows to solve a problem.
Example: A sorting algorithm orders a list of names alphabetically.
Classification
Predicting which category an example belongs to.
Example: Sorting reviews into positive or negative.
Clustering
Grouping similar items together with no labels.
Example: Segmenting customers by behavior.
Embedding
Turning words or items into vectors so similar things sit close.
Example: 'king' and 'queen' land near each other in vector space.
Feature
An input variable the model uses to make a prediction.
Example: Age and income are features for a credit-scoring model.
Gradient Descent
The optimization that nudges a model downhill to lower error.
Example: Each step reduces the loss a little, like rolling into a valley.
Hallucination
When a model states something false with confidence.
Example: An LLM inventing a citation that doesn't exist.
Hyperparameter
A setting you choose before training (e.g. learning rate).
Example: Picking how many layers a network has.
Inference
Using a trained model to make predictions on new input.
Example: Running a photo through a finished classifier.
Label
The correct answer attached to a training example.
Example: Each email is labeled 'spam' or 'not spam'.
LLM
A large language model trained to predict the next token.
Example: Chat assistants are built on LLMs.
MLOps
Practices for deploying and maintaining models in production.
Example: Monitoring a live model for accuracy drift.
Model
A mathematical function learned from data that makes predictions.
Example: A model predicts house prices from size and location.
Neural Network
Layers of connected units loosely inspired by the brain.
Example: Neural networks power face recognition on phones.
Overfitting
When a model memorizes training data and fails on new data.
Example: A model scoring 100% on practice but failing the real test.
Prompt
The instruction you give a generative model.
Example: 'Summarize this report in 3 bullets' is a prompt.
RAG
Retrieval-Augmented Generation: grounding answers in fetched docs.
Example: A support bot quotes your actual help articles.
Regression
Predicting a continuous number rather than a category.
Example: Forecasting next month's revenue.
Temperature
A dial controlling randomness in a model's output.
Example: Low temperature = focused; high = creative.
Token
A chunk of text (word or sub-word) a language model processes.
Example: 'playground' might be two tokens: 'play' + 'ground'.
Training
Adjusting a model's parameters so its predictions match the data.
Example: Training shows the model thousands of labeled photos.
No terms match your search.