GCP Generative AI Leader Practice Question 97: Grounding World Data
Google Cloud Generative AI Leader · Test 10 · question 7 of 10
A travel assistant needs to answer two kinds of questions: what an airline's own baggage policy says, and whether a given airport is currently experiencing weather delays. The baggage policy comes from the airline's own documents, but the weather delay information is general public knowledge the airline does not own. Which grounding source should answer the second kind of question?
Show answer and explanation
Answer:
Current airport weather delays are general, publicly available information the airline does not own and that changes constantly. Grounding with Google Search anchors the model's answers in current public web content, which fits this kind of world knowledge question. The airline's own baggage policy, by contrast, belongs in first-party grounding, for example through Agent Search over the airline's own documents.
Option A is wrong for this question because the airline does not own airport weather data; it fits the baggage policy question instead. Option B is wrong because nothing in the stem describes a licensed third-party data source, only public information. Option C is wrong because relying on what the model learned during training risks a stale or wrong answer about delays that change by the hour; grounding is exactly what avoids that.
Reference: Grounding overview