The model and the big words

Every term in the book, in plain English. Colors match the part each one belongs to.

This page: the big words around AI, and the model. Page 2: the prompt, the harness and the tools.

Artificial intelligence (AI)

Machines doing things we'd call intelligent, by rules people write or by learning.

Machine learning

Getting a computer to learn patterns from data instead of following only hand-written rules.

Deep learning

Machine learning with many layers of simple math units (neural networks).

Generative AI

Models that make new text, images, audio or video.

Large language model (LLM)

A model trained on huge amounts of text to predict the next token. The brain.

Model

The trained program that predicts text. Fixed after training until a new version ships.

Parameters

The internal settings tuned during training. More parameters usually means a bigger, more capable, more expensive model.

Training

Adjusting a model's parameters by having it practice on huge amounts of text, then coaching it with examples and feedback. Happens before you use it.

Pretraining

The first, biggest stage of training: predicting the next word across a huge amount of text.

Transformer

The 2017 design today's language models are built on. It lets every word "look at" every other word at once.

Temperature

A setting for how adventurous the model's word picks are. Low: predictable. High: varied.

Frontier model

One of the largest, most capable models, run in its maker's data centers.

Open-weight model

A model whose trained parameters you can download and run yourself. Not always small.