IBM Assessment: Foundations of AI Intermediate Practice Exam: Medium Difficulty 2025

A retail company wants to implement an AI solution that can analyze customer sentiment from product reviews in multiple languages and automatically route negative feedback to customer service. Which combination of IBM Watson services would be most appropriate for this use case?

A machine learning model deployed in production shows 95% accuracy during training but only 70% accuracy with real-world data. The training dataset contained 10,000 samples collected over one month from a single geographic region. What is the most likely cause of this performance degradation?

An organization is developing an AI system for loan approval decisions. During the ethics review, the team discovers that the historical training data shows approval rates of 75% for one demographic group and 45% for another, reflecting past biased lending practices. What is the most responsible approach to address this issue?

A data science team is comparing supervised learning algorithms for a classification problem. They have a labeled dataset with 50,000 samples across 5 classes. The classes are distributed as follows: Class A (40%), Class B (30%), Class C (15%), Class D (10%), Class E (5%). Which approach should they prioritize to ensure the model performs well across all classes?

A healthcare provider wants to use Watson Discovery to help doctors quickly find relevant medical research papers and patient case studies. The system needs to understand medical terminology and provide evidence-based responses. What key capabilities of Watson Discovery make it suitable for this application?

In the context of AI development, what is the primary distinction between narrow AI (weak AI) and general AI (strong AI), and which type represents current commercial AI applications?

A manufacturing company wants to implement predictive maintenance using machine learning to predict equipment failures before they occur. They have sensor data from machines including temperature, vibration, and pressure readings, along with historical failure records. What type of machine learning problem is this, and what challenges should they anticipate?

An e-commerce company is designing a chatbot using Watson Assistant to handle customer inquiries about order status, returns, and product information. The chatbot needs to hand off complex issues to human agents. What Watson Assistant features should be implemented to ensure effective conversation flow and seamless escalation?

A financial services company is implementing AI for fraud detection and must comply with regulations requiring explainability of automated decisions. Which approach best addresses both the technical requirement for accurate fraud detection and the regulatory requirement for transparency?

A data scientist is training a neural network and observes that the training loss decreases steadily, but the validation loss decreases initially and then starts increasing after epoch 15. The training accuracy is 96% while validation accuracy is 78%. What is occurring, and what is the most appropriate action?