Question

A financial services firm builds an internal gen AI assistant on Vertex AI to answer employee questions about current market conditions and recent regulatory news. During testing, the model gives outdated or fabricated answers because its training data has a fixed cutoff date. The team wants the model's responses to reflect current, verifiable information without retraining the model. Which Vertex AI capability should they use?

Accepted Answer

Grounding the model with Google Search or their own data sources

Answer

Increasing the model's temperature setting to encourage more varied answers

Answer

Expanding the context window to fit more of the training data

Answer

Deploying the model on TPUs to speed up inference