Powerful AI systems that act like advanced autocomplete to understand and write human language. They learn by reading massive amounts of text from across the internet and books.

Large Language Models

The initial stage where an AI reads millions of documents to learn how language works. It identifies patterns, grammar rules, and facts during this massive reading session.

The Training Phase

The process of cutting sentences into small pieces like words or groups of letters. This helps the computer process language as manageable building blocks.

Breaking Down Tokens

A way of turning words into numbers so the computer understands their meaning. Similar words like 'king' and 'queen' are placed close together in this digital map.

Word Maps and Embeddings

The technology that helps the AI focus on the most important words in a sentence. It allows the model to understand that 'it' refers to the 'dog' and not the 'park' in a story.

The Attention Mechanism

How the AI generates text by guessing the most likely next word in a sequence. It does this one word at a time based on all the words that came before it.

Predicting Next Words

The internal 'knobs' or settings that store the AI's knowledge. The more parameters a model has, the more complex patterns and information it can handle.

Billions of Parameters

The specific amount of text the AI can 'remember' during a single conversation. If a chat gets too long, the model may start forgetting the earlier parts.

The Memory Limit

Extra training that teaches the AI to excel at a specific task, like medical advice or coding. It is like taking a graduate student and putting them through law school.

Specialized Fine-Tuning

A process where humans rate the AI's answers to make them more helpful and safe. This helps the model align its behavior with human expectations.

Human Feedback Training

The text inputs you provide to tell the AI what you want it to do. Writing clear prompts helps the AI understand your goal and provide better answers.

Instructions and Prompts

A setting that controls how imaginative or predictable the AI's answers are. A higher setting leads to more creative but sometimes less accurate responses.

The Creativity Dial

Instances where the AI makes up facts that sound very convincing but are actually false. This happens because the AI is a pattern-spotter rather than a database.

What is a Prompt?