Types of AI - ML, DL, Generative AI, and Agents CAIPM Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   Which characteristic is most essential to an AI agent?

   • AThe ability to generate realistic images from text prompts
   • BThe ability to classify data into predefined categories
   • CThe ability to store large amounts of data in a relational database
   • DThe ability to perceive its environment and take actions to achieve goals
   Reveal Solution Discussion Select an answer first
2. 17expert · hard

   A city government wants to optimize traffic light timings to reduce congestion. They have real-time sensor data from intersections, historical traffic patterns, and the ability to change light timings remotely. The system must adapt to unexpected events like accidents and also provide a simulation of proposed changes before deployment. Which combination of AI types is most appropriate?

   • AUse a machine learning model to classify congestion levels and a generative model to adjust timings
   • BUse a deep learning model to predict traffic and a rule-based system to adjust timings
   • CUse a reinforcement learning agent to adjust timings in real time, and a generative model to simulate traffic scenarios
   • DUse an AI agent to simulate traffic and a deep learning model to adjust timings
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A hospital wants to predict patient readmission risk using electronic health records (EHRs), which include both structured data (vitals, lab results) and unstructured clinical notes. The data science team has limited labeled data but access to a large corpus of unlabeled clinical notes. They need a model that is both accurate and interpretable for clinicians. What approach best balances these constraints?

   • ADeploy an AI agent that autonomously reviews each patient's chart and decides readmission risk
   • BUse a generative model to synthesize additional labeled EHR records, then train a simple logistic regression
   • CUse a pre-trained language model to generate embeddings from clinical notes, then train a gradient-boosted tree on the embeddings plus structured features
   • DTrain a deep neural network end-to-end on the raw EHR data
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A healthcare provider wants to deploy an AI system that can triage patient messages and schedule appointments. The system must comply with strict privacy regulations, so it cannot send patient data to external APIs. The provider has a limited IT budget and needs the system to operate reliably with minimal human oversight. Which approach is most appropriate?

   • AUse a deep learning model in the cloud for triage and scheduling.
   • BUse a cloud-based generative AI API for triage and a separate scheduling system.
   • CDeploy an on-premises AI agent that uses a locally hosted model for triage and scheduling.
   • DUse a supervised ML model hosted on-premises for triage, but require human approval for every appointment.
   Reveal Solution Discussion Select an answer first
6. 20expert · hard

   A cybersecurity firm wants to automate threat response. The system must detect anomalies in network traffic, generate incident reports, and take immediate action to isolate affected servers. However, the firm is concerned about false positives causing unnecessary downtime. What approach should they take?

   • AUse an ML model for anomaly detection, a generative model for report drafting, and an AI agent that requires human approval before isolating servers
   • BUse a generative model to create synthetic attack scenarios and train a rule-based system
   • CUse an AI agent that automatically isolates any server with an anomaly score above a threshold
   • DUse a deep learning model to classify traffic and generate reports, but no action
   Reveal Solution Discussion Select an answer first