AI-Based System Testing Fundamentals CT-AI Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   What is the primary purpose of testing input data for an AI system?

   • ATo increase the speed of the model training process
   • BTo ensure the data meets the quality and validity requirements defined for the system
   • CTo replace the need for model validation
   • DTo reduce the size of the training dataset
   Reveal Solution Discussion Select an answer first
2. 2expert · hard

   A healthcare AI system is being developed to diagnose skin lesions from images. The model is intended for use in rural clinics with low-bandwidth internet. The team must balance diagnostic accuracy with the need for fast, offline inference. Which testing strategy best addresses this trade-off?

   • AEvaluate the model's accuracy on a representative test set, and also test a compressed version of the model for speed and accuracy trade-offs.
   • BPrioritize testing the model's robustness to image quality variations, as rural clinics may have poor camera quality.
   • CUse a cloud-based inference service to ensure the model runs on powerful hardware, avoiding the need for compression.
   • DTest the model on a high-performance server to ensure maximum accuracy, then deploy the same model to the clinics.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   What does bias evaluation in ML model testing primarily aim to detect?

   • ASystematic errors that unfairly favor or disadvantage certain groups
   • BThe model's inability to handle large datasets
   • CRandom errors in the training data
   • DThe speed of the model's inference
   Reveal Solution Discussion Select an answer first
5. 4expert · hard

   A company is developing an AI system for autonomous vehicles. The system must be safe (no accidents) and reliable (operate in all weather conditions). The team has limited testing time and budget. Which testing approach is most appropriate?

   • ADefer testing until after deployment and rely on over-the-air updates to fix issues.
   • BPrioritize testing on edge cases and adverse weather conditions, using synthetic data to supplement real-world data.
   • CFocus on testing the model's accuracy on a large test set of normal driving conditions.
   • DTest the model only on real-world data collected from test vehicles.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A team is testing a machine learning model for credit risk assessment. The model's accuracy is 95%, but the team is concerned about bias against a particular ethnic group. What is the most appropriate way to evaluate bias?

   • ACompare the model's false positive and false negative rates across different ethnic groups.
   • BTest the model on a dataset that excludes ethnicity data.
   • CCheck the model's overall accuracy on a test set that includes all ethnic groups.
   • DReview the model's feature importance to see if ethnicity is a significant feature.
   Reveal Solution Discussion Select an answer first