2.4 Explain Defensive Prompting and Mitigation Strategies for AI-generated Errors (e.g., Hallucinations) AI-TECHNICAL-PRACTITIONER Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A user is asking an AI for advice on a complex technical problem. The AI's first response is vague and could be interpreted in multiple ways. The user wants a more precise and actionable answer. Which follow-up prompt is best?

   • APlease provide an answer that is correct.
   • BPlease provide a more detailed answer.
   • CPlease provide an answer that is easy to understand.
   • DPlease provide a step-by-step answer. For each step, state the expected outcome and any assumptions you are making. If you are uncertain about a step, say so.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   What is the primary purpose of defensive prompting when interacting with an AI model?

   • ATo increase the speed at which the AI model generates responses
   • BTo structure prompts in a way that reduces the likelihood of AI-generated errors
   • CTo make the AI model's responses more creative and varied
   • DTo prevent unauthorized access to the AI model's training data
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   Which prompt includes an effective constraint to limit hallucination risk?

   • AWrite an essay about any topic you like.
   • BList three causes of the French Revolution, and only use information from the provided textbook.
   • CTell me about the history of the internet.
   • DDescribe a typical day in the life of a Roman soldier.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A developer is using an AI code assistant to generate a function. The first output contains a bug. Which iterative refinement approach is most likely to produce a correct function?

   • AAsk the assistant to generate the function again from scratch.
   • BProvide the erroneous output, explain the specific bug, and ask the assistant to fix only that issue while keeping the rest unchanged.
   • CAsk the assistant to write a similar function for a different problem.
   • DAccept the output and manually fix the bug later.
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   An AI system is used to answer questions about a company's internal IT policies. Users report that the AI sometimes gives answers that are plausible but not actually in the policy documents. The system is already grounded with the policy documents. What is the most likely remaining cause of these hallucinations?

   • AThe AI is generating responses based on patterns in its training data rather than strictly following the provided documents.
   • BThe users are asking questions that are not covered by the policy documents.
   • CThe AI is intentionally trying to be helpful by providing additional information.
   • DThe policy documents are too long for the AI to process.
   Reveal Solution Discussion Select an answer first