liveBook · Manning

Large language model

As explained in Self-Hosting LLMs at Scale

A machine learning model trained on large text collections to understand and generate language.

A large language model (LLM) is a program trained on enormous amounts of data, mostly text, that can read text and generate a response. Text can include source code and structured data. Some current models also accept images or audio, but this deployment example uses text in and text out. section 1.1

Organizations can send user text to a provider's API or download and run a model on infrastructure they control. Self-hosting means data does not have to leave those systems; a managed cloud service can keep requests within a cloud account while the provider runs the model. section 1.0 section 1.1

Confidential client documents

A law firm discovers that confidential client material entered a public AI tool and reached computers it could neither control nor inspect. Its response is to seek the same capability while keeping its data under its control. section 1.1

How Self-Hosting LLMs at Scale covers it

How deep it goes
Defines an LLM and limits deployment examples to text-in/text-out, while noting image and audio inputs without treating them.
Examples
Uses a law firm’s confidential client material sent to a public AI tool to motivate controlled deployment.
What it assumes
Assumes comfort with Python and the command line, but no prior language-model or GPU experience.
What you can do after
Evaluate whether provider-hosted APIs or self-hosting better fits organizational data-control needs.
How it connects
Links LLMs to APIs, self-hosting, and multimodal models.

Also explained in

Compare how 6 books explain large language model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Self-Hosting LLMs at Scale. The section links point to the parts of the book it draws on.