Generative AI and Large Language Models (LLMs) AIE Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A data scientist is explaining to a manager why an LLM can answer questions about many topics without being explicitly programmed for each one. Which explanation is correct?

   • AThe model stores a copy of the entire training corpus and retrieves exact sentences.
   • BThe model has a built-in internet connection that it uses to look up answers in real time.
   • CThe model was manually programmed with rules for every possible question.
   • DThe model was pre-trained on a vast and diverse text corpus, learning statistical patterns of language and world knowledge.
   Reveal Solution Discussion Select an answer first
2. 27application · medium

   A startup is building a tool that generates realistic product images from text descriptions for an online catalog. Which type of AI model is most appropriate for this task?

   • AA clustering model, because it can group similar product images and then interpolate between them.
   • BA discriminative AI model, because it can classify images into categories and then select one to display.
   • CA generative AI model, because it learns the distribution of the training data and can produce new, plausible images.
   • DA regression model, because it can predict continuous pixel values from the text description.
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A customer support team uses an LLM to generate responses to user emails. They notice that the model sometimes provides overly verbose answers and occasionally includes irrelevant details. They want to improve response quality without retraining the model. Which prompt engineering technique would be most effective?

   • AUse a longer prompt with more background information about the company.
   • BSwitch to a larger LLM that is known to produce shorter responses.
   • CRemove all examples from the prompt to avoid confusing the model.
   • DAdd explicit constraints to the prompt, such as 'Answer in 3 sentences or fewer' and 'Only address the user's question.'
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   Which real-world application commonly uses LLMs?

   • AAutomated language translation services
   • BDatabase index optimization
   • CHardware firewall configuration
   • DCPU scheduling in operating systems
   Reveal Solution Discussion Select an answer first
6. 30expert · hard

   A non-profit organization uses an LLM to generate responses to public inquiries about sensitive topics. They have observed that the model sometimes produces biased or offensive content. They want to reduce this behavior but have limited technical expertise and cannot fine-tune the model. They also need to maintain a respectful tone. Which strategy is most effective?

   • AUse a larger LLM that is known to be unbiased.
   • BDisable the model's ability to generate opinions by using a special setting.
   • CSwitch to a rule-based chatbot that uses pre-approved responses.
   • DCraft prompts that explicitly instruct the model to be neutral, respectful, and to avoid stereotypes, and review outputs periodically.
   Reveal Solution Discussion Select an answer first