1.4 AI/Machine Learning 400-007 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   What is the primary purpose of implementing secure access controls in an AI/ML environment?

   • ATo ensure only authorized personnel can access models, data, and training pipelines
   • BTo reduce the latency of inference requests
   • CTo simplify the deployment of AI/ML models
   • DTo improve the accuracy of model predictions
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   What is the primary purpose of an AI/ML governance framework?

   • ATo maximize the performance of AI/ML models
   • BTo automate the deployment of AI/ML models
   • CTo reduce the cost of AI/ML infrastructure
   • DTo establish policies and processes that ensure AI/ML systems are ethical, compliant, and well-managed
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A healthcare organization is deploying an AI model to assist in diagnosing medical images. The model is trained on sensitive patient data. The organization must ensure that only authorized radiologists can access the model's inference API and that the model's predictions cannot be tampered with. Which combination of controls should the organization implement?

   • AGrant all clinicians access to the inference API and log all requests for auditing.
   • BRely on network segmentation to isolate the AI service, and use plaintext HTTP for internal communication.
   • CUse API keys for authentication and store the model in a public repository for transparency.
   • DImplement role-based access control (RBAC) with multi-factor authentication (MFA) for the inference API, and sign the model artifacts to ensure integrity.
   Reveal Solution Discussion Select an answer first
5. 9expert · hard

   A multinational corporation is designing an AI/ML platform for its global operations. The platform must comply with data residency laws in multiple countries, including the EU and China. The company also wants to minimize costs by avoiding data duplication. Which design strategy best balances compliance and cost?

   • AStore all data in the cloud provider's headquarters country and rely on legal agreements to ensure compliance.
   • BDeploy a single global AI platform in a region that is acceptable to all regulators, such as a neutral offshore location.
   • CUse a hybrid cloud with a single data lake in the US and replicate only metadata to other regions.
   • DUse a federated learning approach where models are trained locally in each region and only model updates are shared globally.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   An e-commerce company has deployed a recommendation engine that uses real-time user behavior data. The model's performance has degraded over time, and the company wants to detect when the model's predictions become unreliable. Which approach should the company implement to maintain trustworthy outcomes?

   • ARetrain the model weekly without any monitoring, assuming that regular retraining ensures reliability.
   • BDisable the model and use a rule-based system to ensure consistent outcomes.
   • CSet up automated monitoring of model accuracy against a holdout dataset and alert when accuracy drops below a threshold.
   • DRely on user feedback to identify model issues, since users will report poor recommendations.
   Reveal Solution Discussion Select an answer first