Page 38 - CyberSecurityEssentialsEnglish

Models can summarize text, translate, write computer code, and answer questions because
they have learned many linguistic patterns. Yet constructing a suitable sentence is different
from finding a current fact or performing an exact calculation. A task may therefore require

a reliable source, current search, calculator, or database, in addition to human review.

✕ Common misconception Correction

The model retrieves the correct sentence from The model generates a probable sequence
an internal memory like a database, and based on the context and what it learned. The

confident writing means it knows the answer. result may match the facts or express a
convincing error, especially when a source is
missing or the information is recent.

⚠ Warning: Linguistic completion can invent a reference
If a model is asked for the title of a study or the number of a legal provision without
access to search or a source, it may produce a plausible title, author, and number that do

not exist. A well-formatted reference is not enough; its existence and accuracy must be
verified.

Video: Introduction to Large Language Models
An introduction to large language models,

pretraining, tokens, context, and the use of
one model for multiple tasks.

Duration: about 12 minutes | Level: Beginner
to Intermediate
Learning objective: Explain how a language
model predicts text and why it differs from a

search engine and a database.
Source: Google Cloud Tech [15].
https://youtu.be/zizonToFXDs

Verified: June 6, 2026

“Fluent language tells you that a sentence is probable and coherent. By itself, it does not
tell you that the sentence is true, current, or appropriate for your decision.”

● Remember: Section summary
• Key idea: A language model predicts the next unit of text, so fluency is evidence of
probability, not of truth.

37