Task Statement 4.1: Explain the Development of AI Systems That Are Responsible. AIF-C01 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   A machine learning team wants to detect bias in their model by analyzing its performance across different demographic subgroups. Which AWS service is specifically designed for this purpose?

   • AAmazon SageMaker Clarify
   • BAmazon SageMaker Model Monitor
   • CAmazon Augmented AI (A2I)
   • DAmazon Bedrock Guardrails
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   A generative AI application provides medical advice to users. The model occasionally produces confident but incorrect information that could lead a user to take harmful action. Which legal risk is most directly associated with this scenario?

   • AHallucinations
   • BIntellectual property infringement
   • CBiased outputs
   • DLoss of customer trust
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A government agency is building a resume-screening AI to shortlist candidates for public sector jobs. The initial dataset contains resumes from a 10-year period when the agency hired predominantly from two local universities. The data science team is evaluating the dataset for responsible AI readiness. Which dataset characteristic is most likely to be problematic for this use case?

   • ALack of curation, because the resumes were not manually reviewed by HR professionals before inclusion
   • BLack of balance, because the dataset contains an unequal number of resumes from each of the two universities
   • CLack of inclusivity, because the dataset does not include candidates with disabilities or from minority groups
   • DLack of diversity, because the dataset does not represent the broader population of qualified candidates
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A startup is choosing a foundation model for a text summarization feature in their mobile app. The team has identified two models with comparable summarization quality. Model A requires 4 GB of GPU memory for inference, while Model B requires 16 GB. The startup runs on a limited budget and wants to minimize operational costs while maintaining quality. Which consideration should be the primary factor in the model selection decision?

   • AThe robustness of the models, since Model B may be more resilient to adversarial inputs
   • BThe inclusivity of the models, since Model B may handle a wider range of languages and dialects
   • CThe environmental impact and sustainability of the models, since Model A will consume fewer compute resources
   • DThe veracity of the models, since Model B may produce more factual summaries
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A financial services company is deploying a generative AI chatbot on Amazon Bedrock to answer customer questions about loan products. The compliance team is concerned about the chatbot providing specific, unsupported interest rate quotes that could constitute binding offers. The company wants to enforce a policy that the chatbot must decline to answer rate questions and instead direct customers to a human agent. Which Amazon Bedrock Guardrails configuration should the team implement?

   • AConfigure a PII redaction filter to remove any numeric values that resemble interest rates from the output
   • BConfigure a prompt injection filter to prevent users from overriding the system prompt about rates
   • CConfigure a denied topic for 'loan interest rates' with a configured message that routes customers to a human agent
   • DConfigure a word filter to block the terms 'interest rate' and 'APR' from all model inputs and outputs
   Reveal Solution Discussion Select an answer first