Challenges Testing Complex AI-Based Systems CT-AI Practice Questions (Page 6)

Questions 26–27

1. 26foundation · easy

   Why does the complexity of deep neural networks make it difficult to predict their behavior?

   • AThey are always non-deterministic, so behavior cannot be predicted.
   • BThey are too small to be tested effectively.
   • CThey do not use any mathematical functions.
   • DThe large number of parameters and non-linear transformations create a highly complex function that is hard to analyze.
   Reveal Solution Discussion Select an answer first
2. 27expert · hard

   A team is testing a facial recognition system for a security application. The system must work across different ethnicities, ages, and lighting conditions. The team has a limited budget and cannot collect a perfectly balanced dataset. What is the most effective strategy to maximize test coverage given the constraints?

   • ATest only on the most common demographic to ensure the system works for the majority of users.
   • BUse synthetic data to generate a perfectly balanced dataset at no additional cost.
   • CCollect as much data as possible from easily available sources, even if it is mostly from one demographic.
   • DPrioritize collecting test data for underrepresented groups and edge-case lighting conditions, even if it means reducing the overall dataset size.
   Reveal Solution Discussion Select an answer first