AI Models, Frameworks, Strategies, and Use Cases AAIR Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   During which stage of the AI model lifecycle does an organization typically validate that the model meets performance and fairness criteria before production use?

   • ARetirement
   • BValidation
   • CDeployment
   • DMonitoring
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A media company wants to use AI to personalize content recommendations. They have a large user base and significant data science expertise. Their main business goal is to increase user engagement, and they are willing to accept moderate risk. Which strategy is most appropriate?

   • AUse a generic open-source recommendation model without customization
   • BBuild a custom recommendation model using their proprietary user data
   • CBuy a third-party recommendation engine and integrate it with their platform
   • DOutsource the development to a third-party vendor and retain no internal oversight
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   What is the primary purpose of the ONNX (Open Neural Network Exchange) format in the AI ecosystem?

   • ATo automate the entire model training process without human intervention.
   • BTo replace all other AI frameworks as the single official development tool.
   • CTo provide a standardized format for representing trained models so they can be moved between different frameworks.
   • DTo enforce ethical guidelines for AI model deployment.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A manufacturing company has developed a predictive maintenance model using TensorFlow. The model performs well in offline testing, but the team is unsure how to deploy it to their on-premises edge devices. Which framework or approach would best facilitate deployment while maintaining governance controls?

   • ARetrain the model using PyTorch to ensure compatibility with edge devices
   • BConvert the model to ONNX format and use an ONNX runtime on the edge devices
   • CUse TensorFlow Serving on each edge device
   • DDeploy the model as a cloud service and have edge devices send data to it
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which AI model category is best suited for generating new, previously unseen content such as realistic images or text based on learned patterns?

   • AClassical AI (expert systems)
   • BSupervised machine learning
   • CReinforcement learning
   • DGenerative AI
   Reveal Solution Discussion Select an answer first