3.1 Describe How to Proactively Overcome Foundation Model Limitations. GENERATIVE-AI-LEADER Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   A healthcare organization is building a patient-facing chatbot to answer questions about prescription medications. The model was trained on general web data and sometimes provides outdated or incorrect dosage information. The team needs the chatbot to always reference the latest FDA-approved drug labels. Which approach best addresses this limitation?

   • AAdd a disclaimer that the chatbot may be inaccurate and should not be used for medical decisions.
   • BUse a larger foundation model with more parameters to reduce the chance of hallucination.
   • CImplement retrieval-augmented generation (RAG) with the FDA drug label database as the grounding source.
   • DFine-tune the foundation model on historical patient chat logs to improve conversational tone.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   A user asks a foundation model to summarize a legal document. The model produces a summary that includes a clause that does not exist in the original document. Which foundation model limitation is most directly demonstrated?

   • AKnowledge cutoff
   • BEdge case
   • CBias
   • DHallucination
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A company has a generative AI model that generates personalized email responses. The model was performing well, but after a major platform update, the team notices the responses have become less coherent. The model version and configuration are unchanged. What is the most likely cause and the best first action?

   • AThe model has reached its knowledge cutoff; switch to a newer model.
   • BThe temperature setting has been corrupted; reset the configuration.
   • CThe input data has drifted; immediately retrain the model on new email data.
   • DThe underlying model API may have been updated; check the model version and review the change logs.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A global e-commerce company is using a generative AI model to write product reviews summaries. The model performs well for English reviews but produces poor summaries for reviews in other languages. The team has a small set of translated example summaries. What is the most effective way to improve multilingual performance?

   • AUse few-shot prompting with translated examples for the underperforming languages.
   • BUse a larger model that is known to support more languages.
   • CTranslate all reviews to English before summarizing them.
   • DFine-tune the model on the small set of translated examples.
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A company has a generative AI model that generates marketing copy. The model is accessed via an API that the provider updates automatically. The team has noticed a gradual decline in output quality over several weeks. They need to maintain consistent quality. What is the most appropriate strategy?

   • APin the model version and set up a monitoring dashboard to track quality metrics before considering upgrades.
   • BFine-tune the model every week on the latest marketing copy to keep it current.
   • CDisable the model and use a rule-based template system for all marketing copy.
   • DAllow automatic upgrades and rely on the provider's quality guarantees.
   Reveal Solution Discussion Select an answer first