Types of AI - ML, DL, Generative AI, and Agents CAIPM Practice Questions (Page 7)

Questions 31–35

1. 31application · medium

   A retail company wants to automatically categorize customer support tickets into predefined topics (e.g., billing, shipping, product issue) based on the ticket text. The team has a labeled dataset of 10,000 historical tickets. They need a solution that is interpretable and can be retrained quickly as new topics emerge. Which approach best meets these requirements?

   • AApply an unsupervised clustering algorithm to group tickets and then manually assign labels to each cluster.
   • BUse a deep learning model with a custom architecture trained from scratch on the raw text.
   • CTrain a supervised machine learning classifier such as logistic regression on the labeled ticket text.
   • DDeploy a generative AI model to generate new ticket categories from the text.
   Reveal Solution Discussion Select an answer first
2. 32application · medium

   A marketing team needs to produce personalized product descriptions for thousands of items across multiple languages. They have a catalog with product attributes and past examples of high-performing copy. What should they implement?

   • AAn AI agent that autonomously updates the entire e-commerce website
   • BA deep learning classifier that categorizes products by type
   • CA predictive ML model that scores each existing description for engagement
   • DA generative AI model fine-tuned on the existing copy examples
   Reveal Solution Discussion Select an answer first
4. 33foundation · easy

   Which statement best defines machine learning in a business context?

   • AA technique that generates entirely new data samples that resemble a training dataset
   • BA set of rules explicitly programmed by developers to produce outputs from inputs
   • CA statistical approach that enables systems to improve performance on a task by learning patterns from data
   • DA method that uses neural networks with many layers to automatically extract hierarchical features
   Reveal Solution Discussion Select an answer first
5. 34expert · hard

   A government agency wants to analyze satellite imagery to detect illegal deforestation. They have a small set of labeled images from experts, but millions of unlabeled images. They need high accuracy and the ability to explain why a particular area was flagged. What approach should they take?

   • AUse a generative model to create synthetic labeled images, then train a simple logistic regression on pixel values
   • BDeploy an AI agent that autonomously flags areas based on a set of hand-coded rules
   • CUse self-supervised deep learning to pre-train on unlabeled images, then fine-tune on labeled images, and use saliency maps for explanations
   • DTrain a deep learning model from scratch on the labeled images only
   Reveal Solution Discussion Select an answer first
6. 35application · medium

   A retail chain wants to predict which customers are likely to churn next quarter based on historical purchase data, demographics, and support interactions. The data is structured, moderately sized, and the team needs a model that can be easily interpreted by business stakeholders. Which approach should the team choose?

   • AUse a generative model to synthesize new customer profiles and then cluster them
   • BTrain a gradient-boosted decision tree on the labeled historical data
   • CDeploy a deep neural network to automatically learn complex patterns from raw transaction logs
   • DImplement a rule-based system that flags customers with low recent activity
   Reveal Solution Discussion Select an answer first