AI-Based System Testing Fundamentals CT-AI Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A team is testing an AI system for sentiment analysis. The test data includes text from social media, product reviews, and news articles. What is the best way to manage this test data?

   • ACreate separate test sets for each text type and evaluate the model's performance on each.
   • BCombine all text types into a single test set and evaluate the model's overall performance.
   • CUse only the social media data because it is the most representative of real-world usage.
   • DRemove the news articles because they are too formal for sentiment analysis.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which of the following is a core component that should be explicitly defined in the specification of an AI-based system?

   • AThe performance requirements, such as accuracy or response time
   • BThe names of the development team members
   • CThe marketing strategy for the AI product
   • DThe exact source code of the machine learning algorithm
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   An AI system is designed to detect fraudulent credit card transactions. The model was trained on a highly imbalanced dataset where only 0.1% of transactions are fraudulent. During testing, the model achieves 99.9% accuracy. Which additional evaluation metric is most important to assess?

   • AMean absolute error (MAE)
   • BPrecision and recall for the fraudulent class
   • CF1 score for the non-fraudulent class
   • DR-squared
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   Which of the following is a method used to evaluate the robustness of a machine learning model?

   • ACounting the number of features used
   • BReviewing the model's architecture diagram
   • CTesting the model with adversarial or perturbed inputs
   • DChecking the model's training time
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A company is developing an AI system to detect cyberattacks in real-time. The system must have low latency and high accuracy. The team has a small labeled dataset of attacks, but a large amount of unlabeled network traffic data. Which testing approach is most appropriate?

   • ADeploy the model immediately and monitor its performance in production, as real-time data is the best test.
   • BTest the model only on the labeled dataset to ensure accuracy, and accept the latency as is.
   • CUse the unlabeled data for semi-supervised learning to improve the model, then test on a held-out labeled set, and also measure inference latency.
   • DFocus on testing the model's robustness to adversarial attacks, as this is more important than accuracy.
   Reveal Solution Discussion Select an answer first