5.3 Explain AI Workflow Design and Monitoring Principles AI-TECHNICAL-PRACTITIONER Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A telecommunications company has deployed a churn prediction model. The model's performance has been declining over time. The data team notices that the distribution of customer usage patterns has changed. What should the team do?

   • ARevert to an older version of the model that performed well in the past.
   • BIgnore the change because the model was validated on historical data.
   • CIncrease the model's complexity to capture the new patterns.
   • DRetrain the model with recent data that reflects the new usage patterns.
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A multinational corporation is deploying an AI system for employee performance evaluation. The system uses multiple data sources, including productivity metrics and peer reviews. The company must comply with data protection regulations in different countries, some of which restrict cross-border data transfers. The AI team wants to use a centralized cloud platform for model training. What should the company do?

   • AAnonymize the data before transferring it to a central location for training.
   • BTransfer all data to a single cloud region and rely on the cloud provider's compliance certifications.
   • CUse a hybrid approach where data is processed locally and only model updates are shared centrally.
   • DUse a cloud region located in each country to process and store data locally, and train the model in each region.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   An e-commerce company has deployed a recommendation model. They notice that the click-through rate has been declining over the past month, but the model's accuracy on the test set remains high. The data engineering team confirms that the data pipeline is running without errors. What should the team investigate first?

   • ACheck for data drift by comparing the distribution of recent input features with the training data distribution.
   • BIncrease the frequency of model retraining to daily to keep the model fresh.
   • CRoll back to the previous version of the model to restore performance.
   • DRetrain the model immediately with the most recent data to improve performance.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   What is a common source of feedback data for improving a deployed AI model?

   • AUser corrections or ratings of model outputs
   • BThe model's initial training code
   • CThe model's deployment date
   • DThe hardware specifications of the server
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A bank is deploying an AI model to approve loan applications. The model is trained on historical data that includes a biased sample. The bank wants to ensure fairness and transparency in the model's decisions. What should the bank do?

   • AKeep the model's logic confidential to protect the bank's competitive advantage.
   • BAudit the model for bias across protected groups and document the model's decision-making process.
   • CUse a more complex model to reduce bias automatically.
   • DRely solely on the model's accuracy metrics to ensure it is fair.
   Reveal Solution Discussion Select an answer first