By skyforbes A beginner’s visual guide to the Transformer magic behind ChatGPT
When you chat with ChatGPT, it feels like magic. You type something and instantly get a thoughtful, fluent response. But what’s really happening inside? Let’s break it down in simple terms.
ChatGPT Is a Type of “Language Model”
ChatGPT is a Large Language Model (LLM), i.e., a computer program trained to predict the next word in a sentence.
It learns this by reading billions of sentences from books, articles, and the internet. Every time, it tries to guess what word comes next and slowly, it gets really, really good at it.
Step 1: Text Becomes “Tokens”
Before ChatGPT can understand your message, it breaks your text into tiny chunks called tokens — roughly, word pieces. A token can be a word, part of a word, or even punctuation, depending on the model’s vocabulary.
For example:
“ChatGPT works great” → [“Chat”, “G”, “PT”, “works”, “great”]
These tokens are converted into numbers (vectors) so the model can do math with them.