liveBook · Manning

Factuality

As explained in Introduction to Generative AI, Second Edition

Factuality is the degree to which generated content accurately represents facts about the world.

A legal brief can sound convincing while citing lawsuits that never happened. Factuality is the degree to which generated content accurately represents facts about the world. Plausible writing is not, by itself, evidence that a claim is true. section 5.8

Large language models (LLMs) can summarize information and produce human-like text, but they generally cannot ensure factuality or accuracy. Their fluent responses can lead users to accept claims without checking them against reliable sources. Verification is especially important in fields requiring specialized knowledge, such as law, medicine, and finance. section 5.8

Mata v. Avianca

An attorney used ChatGPT to help compile a legal brief. It invented supporting arguments based on lawsuits that had never occurred. The attorney said he had not intended to deceive the court and had been unaware that the tool’s content could be false. section 5.8

How Introduction to Generative AI, Second Edition covers it

How deep it goes
Explains the limitation through a legal case but does not provide a factuality measurement method.
Examples
Uses a legal brief containing invented lawsuits to illustrate false generated claims.
What you can do after
Evaluate generated claims by checking them against reliable sources.
How it connects
Connects factuality to large language models, text summarization, and verification against reliable sources.

Also explained in

Compare how 2 books explain factuality →

Related concepts

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