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.

Factuality is the degree to which generated content accurately represents facts about the world. A large language model can produce convincing prose without ensuring that its claims are true. That difference matters when people use its output for professional work. section 5.8

An invented legal precedent

In Mata v. Avianca, an attorney used ChatGPT to prepare a brief containing references to lawsuits that had never occurred. He later told the judge that he had not realized the tool could produce false content. The plausible-looking brief did not make its supporting claims real. section 5.8

Check consequential claims against reliable sources before using them. This is especially important in fields requiring specialized knowledge, including law, medicine, and finance. Clearer expectations about what these tools can and cannot verify also matter. section 5.8

How Introduction to Generative AI, Second Edition covers it

How deep it goes
Distinguishes plausible prose from accurate claims, but gives no factuality score or formal testing procedure.
Examples
Uses a ChatGPT-assisted legal brief citing nonexistent lawsuits in Mata v. Avianca.
What you can do after
Evaluate/understand it by checking plausible generated claims against reliable sources before relying on them professionally.
How it connects
Links factuality to large language models, source verification, and professional work in law, medicine, and finance.

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.