liveBook · Manning

Language model

As explained in AI Agents in Action

A large language model is a neural language model trained on extensive text data to generate and transform natural-language content.

A large language model (LLM) is a generative model trained on extensive text data to create natural-language content. Unlike predictive or classification models, it generates content rather than simply assigning a class. section 2.0 section 2.6

Figure 2.1 The difference between generative and predictive models
figure

Compare how generative models create content with how predictive and classification models classify it. section 2.0

An LLM combines training data, model architecture, and training for a particular use case. Fine-tuning further adapts a model to a domain or task; GPT models use transformer architecture and can scale to billions of parameters. section 2.0 section 2.6

Figure 2.2 The main elements that describe an LLM
figure

Use the diagram to see the main elements that describe an LLM. section 2.0

ChatGPT as a chat-completions model

ChatGPT is trained on public internet data and refined through training strategies, including reinforcement learning with human feedback, to support chat completions. section 2.0

LLMs can power chatbots, assistants, and agents. They can be accessed through APIs or explored locally with tools such as LM Studio; prompt engineering helps shape more effective requests. section 2.6

How AI Agents in Action covers it

How deep it goes
Defines data, architecture, use-case training, and fine-tuning, and describes transformer scaling to billions of parameters.
Examples
Illustrates chat completions with ChatGPT and describes API access and local use with LM Studio.
What it assumes
Assumes a Python virtual environment and access to GPT-4 or above for the Prompt Flow connection exercise.
What you can do after
Apply language models through the OpenAI API SDK or local tools such as LM Studio.
How it connects
Connects language models with GPT, transformer architecture, fine-tuning, reinforcement learning with human feedback, and prompt engineering.

Also explained in

Compare how 69 books explain language model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Agents in Action. The section links point to the parts of the book it draws on.