liveBook · Manning

Language model

As explained in Deep Learning with PyTorch, Second Edition

A language model with enough scale to generate and answer with fluent text from prompts.

A large language model can take a prompt and produce fluent text, including answers to questions. ChatGPT is a chatbot that uses one. Its responses can be strikingly coherent, but that does not mean the machine thinks or is self-aware in a human sense. section 1.1

A prompt and a response

Prompted to introduce itself through a poem, ChatGPT produced several rhyming lines about assisting and conversing. This is text generated in response to an instruction, not a poem supplied as part of the prompt. section 1.1

The ability to perform this kind of task is learned from examples rather than written as a set of handcrafted rules. Language-model output is determined probabilistically, so the same prompt may not always produce the same text. Building a language model also involves the broader deep-learning workflow: representing data, defining a neural network, and training it. section 1.1 section 1.0

How Deep Learning with PyTorch, Second Edition covers it

How deep it goes
Demonstrates prompting and notes probabilistic output, but does not explain language-model architecture or generation mechanics.
Examples
Uses a request for a self-introduction poem and ChatGPT’s generated verse.
What it assumes
Assumes Python programming experience for the hands-on deep-learning workflow, while introducing its concepts from the basics.
What you can do after
Evaluate/understand prompt-driven text generation without treating fluent output as evidence of human-like thought.
How it connects
Connects language models to deep learning, neural networks, prompts, and learning from examples.

Also explained in

Compare how 67 books explain language model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with PyTorch, Second Edition. The section links point to the parts of the book it draws on.