1.4 AI/Machine Learning 400-007 Practice Questions (Page 6)

Questions 26–30

1. 26application · medium

   A government agency is deploying an AI system to automate the approval of public benefits applications. The agency must ensure that the system does not discriminate against any protected group and that decisions can be explained to applicants. Which governance practice is most critical to implement?

   • AAllow the model to make final decisions without human review to ensure efficiency.
   • BStore all training data indefinitely to ensure full traceability.
   • CConduct a bias audit on the training data and model outputs, and document the results.
   • DUse a black-box model to maximize accuracy, and keep the model details confidential.
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   An online gaming company runs an AI-based anti-cheat system that analyzes player behavior in real time. The system must scale to handle sudden spikes in player activity during game launches and events. The company has a limited budget and wants to avoid over-provisioning. The system also needs to maintain low latency for real-time analysis. Which scaling strategy should the company use?

   • AProvision a fixed number of instances that can handle the maximum expected load.
   • BUse a predictive auto-scaling policy that analyzes historical player activity patterns and pre-provisions capacity before expected spikes.
   • CUse a reactive auto-scaling policy based on CPU utilization, with a cooldown period to avoid thrashing.
   • DUse a serverless function that scales to zero when idle, accepting cold starts during spikes.
   Reveal Solution Discussion Select an answer first
4. 28foundation · easy

   Which security concern is specific to AI/ML systems and not typically a focus for traditional IT workloads?

   • AModel poisoning attacks that alter training data to degrade model performance
   • BUnauthorized access to application servers
   • CNetwork intrusion through open firewall ports
   • DData loss due to hardware failure
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   Which deployment model is most appropriate when an organization must keep AI/ML training data within a specific country but wants to use cloud-based compute for burst processing?

   • APrivate cloud only
   • BHybrid cloud
   • CCommunity cloud
   • DPublic cloud only
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A financial services firm is building a credit risk model using data from multiple internal systems. The data sources have inconsistent formats and some records contain missing values. The firm's compliance team requires that all data used in the model be traceable to its origin. What should the firm do to ensure data integrity throughout the AI/ML lifecycle?

   • AUse only the most recent data snapshot and discard historical data to avoid inconsistencies.
   • BUse a data lake to store raw data and rely on the model to handle missing values automatically.
   • CClean the data manually in spreadsheets before loading it into the training system.
   • DImplement a data pipeline that standardizes formats, imputes missing values, and maintains a data lineage record for every record.
   Reveal Solution Discussion Select an answer first