Microsoft Azure AI Fundamentals (AI-900) 4 â Questions and Answers
Question 1: Which type of AI workload involves training a model on labeled examples and using it to predict the label for new, unseen data?
- Unsupervised learning
- Reinforcement learning
- Supervised learning (Correct answer)
- Self-supervised learning
Correct answer: Supervised learning
Supervised learning uses labeled training data so the model can learn a mapping from inputs to outputs and generalize to new examples.
Question 2: In Azure Cognitive Services Language, what does the 'key phrase extraction' feature return?
- A list of named entities like people and places
- The most important phrases that summarize the content of a document (Correct answer)
- A sentiment polarity score for the document
- A translation of the document into another language
Correct answer: The most important phrases that summarize the content of a document
Key phrase extraction identifies and returns the main talking points of a text, useful for quickly summarizing large volumes of content.
Question 3: What is 'overfitting' in machine learning?
- The model performs poorly on both training and test data
- The model learns the training data too well and generalizes poorly to new data (Correct answer)
- The model is trained with too few features to be useful
- The model's predictions are biased toward the majority class
Correct answer: The model learns the training data too well and generalizes poorly to new data
An overfitted model memorizes training data noise and details, causing it to perform worse on unseen data than on the data it was trained on.
Question 4: Which Azure service enables you to add a conversational AI interfaceâsuch as a chatbotâto websites, apps, or messaging channels?
- Azure Logic Apps
- Azure Bot Service (Correct answer)
- Azure Event Grid
- Azure API Management
Correct answer: Azure Bot Service
Azure Bot Service provides a managed platform for building, testing, deploying, and connecting intelligent bots to multiple communication channels.
Question 5: What is the purpose of 'cross-validation' when evaluating a machine learning model?
- To test the model against data from a different country or language
- To estimate model performance more robustly by training and evaluating on multiple data splits (Correct answer)
- To compare two different algorithms on the same hardware
- To ensure the model passes regulatory compliance checks
Correct answer: To estimate model performance more robustly by training and evaluating on multiple data splits
Cross-validation partitions data into k folds, trains on k-1 folds, and tests on the remaining fold repeatedly, giving a more reliable performance estimate.
Question 6: Which Responsible AI principle requires that AI systems be explainable so that humans can understand the decisions they make?
- Accountability
- Fairness
- Transparency (Correct answer)
- Inclusiveness
Correct answer: Transparency
The Transparency principle demands that AI systems and their decisions be understandable to stakeholders, enabling informed oversight and trust.
Question 7: A company uses Azure Cognitive Search to build an internal knowledge base. What AI enrichment capability can automatically extract entities from documents during indexing?
- Semantic ranking
- AI enrichment skills (Skillsets) (Correct answer)
- Synonym maps
- Scoring profiles
Correct answer: AI enrichment skills (Skillsets)
AI enrichment skillsets in Azure Cognitive Search apply cognitive skillsâsuch as entity recognition and key phrase extractionâto documents during the indexing pipeline.
Microsoft Azure AI Fundamentals (AI-900) 4
Which type of AI workload involves training a model on labeled examples and using it to predict the label for new, unseen data?