Module 9 Chapter 3

The Honest Limits

A wrong booking address can make an invitation useless. It can also look perfectly ordinary, tucked into a polished sentence beside the correct time and place.

Language models generate text using learned patterns and supplied information. The resulting sentence may be accurate, unsupported, or partly right. Fluency alone cannot tell you which.

Evidence gives you a way to decide. Find the source, inspect whether it supports the particular claim, and check that it applies to the situation in front of you. A real event page does not support a repair guarantee that the page never makes.

The draft's assumptions deserve attention too. An invitation that assumes everyone can use a smartphone app may exclude some of the people it should welcome. Training material, training decisions, and product choices can all influence the defaults a model produces.

The amount of checking depends on what a mistake would cost. You can discard a weak title in seconds. A published time, booking instruction, or safety promise needs a firmer basis.

These tools can save useful work while leaving important decisions with you. The skill is to recognise what the answer needs before it can be used, then obtain that evidence.

Module 9 Chapter 3

The Honest Limits

A wrong booking address can make an invitation useless. It can also look perfectly ordinary, tucked into a polished sentence beside the correct time and place.

Language models generate text using learned patterns and supplied information. The resulting sentence may be accurate, unsupported, or partly right. Fluency alone cannot tell you which.

The amount of checking depends on what a mistake would cost. You can discard a weak title in seconds. A published time, booking instruction, or safety promise needs a firmer basis.

These tools can save useful work while leaving important decisions with you. The skill is to recognise what the answer needs before it can be used, then obtain that evidence.