Types of AI - ML, DL, Generative AI, and Agents CAIPM Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A legal firm wants to automate the drafting of standard contracts and also review incoming contracts for risky clauses. They have a large library of past contracts and a list of clauses that are considered high-risk. What should they implement?

   • AUse an AI agent to draft contracts and a generative model to review them
   • BUse a generative AI model to draft contracts and a machine learning classifier to flag risky clauses
   • CUse a rule-based system for both drafting and reviewing
   • DUse a deep learning model to generate contracts and an AI agent to manually review them
   Reveal Solution Discussion Select an answer first
2. 12expert · hard

   A cybersecurity firm wants to deploy an AI system that can detect and respond to network intrusions in real time. The system must minimize false positives to avoid alert fatigue, and it must also explain its actions to security analysts. Which approach is most appropriate?

   • AUse a deep learning model alone to detect intrusions and automatically block all suspicious traffic.
   • BUse a supervised ML model for intrusion detection with low false-positive tuning, and an AI agent that takes predefined response actions with logging.
   • CUse a generative AI model to generate alerts for every anomaly.
   • DUse an AI agent that autonomously shuts down the network upon any anomaly.
   Reveal Solution Discussion Select an answer first
4. 13expert · hard

   A credit card company wants to detect fraudulent transactions in real time. The data is highly imbalanced (few fraud cases) and includes both structured transaction features and sequences of user behavior. The team must minimize false positives to avoid annoying customers, but also catch as much fraud as possible. Which approach is most appropriate?

   • AUse an ensemble of gradient-boosted trees with class weighting and anomaly detection features
   • BDeploy an AI agent that automatically blocks any transaction with a risk score above a fixed threshold
   • CUse a generative model to oversample the minority class, then train a simple decision tree
   • DTrain a deep learning model on raw transaction sequences without any feature engineering
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A bank wants to detect fraudulent transactions in real time. They have a labeled dataset of past transactions and need a model that can score each transaction quickly and flag suspicious ones for review. Which approach is most appropriate?

   • AUse a deep learning model that requires extensive training time and computational resources.
   • BUse a generative AI model to create synthetic transactions for analysis.
   • CDeploy an AI agent that autonomously blocks transactions without human review.
   • DTrain a supervised ML model (e.g., logistic regression) on transaction features to output a fraud probability score.
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   An autonomous vehicle company needs to recognize pedestrians, traffic signs, and other vehicles from camera feeds in real time. The system must handle varying weather and lighting conditions. Which AI approach is most suitable?

   • AA traditional ML model using manually engineered features like color histograms
   • BAn AI agent that decides when to brake based on sensor rules
   • CA deep learning model using convolutional neural networks
   • DA generative AI model that creates synthetic driving scenes
   Reveal Solution Discussion Select an answer first