What is a large language model (LLM)?
A.A model that stores every web page it was trained on
LLMs learn patterns, not verbatim storage — that's why they can produce false 'facts'.
B.A deep learning model trained on massive text corpora that generates human-like language by predicting the next tokens
Trained text-prediction models. Correct.
C.A rule-based chatbot engine
Rule-based bots predate LLMs — rigid scripts, not learned language.
D.A database of question-answer pairs
Q&A databases are retrieval — LLMs generate novel text.
Why this answer
LLMs (GPT, LLaMA, etc.) are transformer neural networks trained on vast text — they don't retrieve stored text but generate completions by predicting probable next tokens. This generative mechanism explains both their fluency AND hallucinations.
Correct answer: B · Last reviewed 2026-02-14
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Generative AI practice session.
Practice Generative AI