Microsoft Azure AI Fundamentals (AI-900) Practice Questions

This document contains 25 practice questions covering all domains of the AI-900 exam. Use these questions to test your knowledge and identify areas where you might need additional study.

Domain 1: Describe Artificial Intelligence Workloads and Considerations

Question 1

Which of the following is NOT a common AI workload?

  • A) Computer vision
  • B) Natural language processing
  • C) Database management
  • D) Generative AI

Answer: C) Database management

Explanation: Database management is a traditional IT workload, not specifically an AI workload. Computer vision, natural language processing, and generative AI are all common AI workloads covered in the AI-900 exam.

Question 2

Which principle of responsible AI focuses on ensuring AI systems treat all people fairly?

  • A) Transparency
  • B) Fairness
  • C) Accountability
  • D) Privacy and security

Answer: B) Fairness

Explanation: Fairness in responsible AI ensures that AI systems treat all people fairly and avoid bias in data and algorithms. Transparency relates to making AI systems understandable, accountability relates to taking responsibility for AI systems, and privacy and security relate to protecting user data.

Question 3

A company wants to implement AI in their healthcare system. Which responsible AI principle should they prioritize to ensure patient data is protected?

  • A) Inclusiveness
  • B) Reliability and safety
  • C) Privacy and security
  • D) Transparency

Answer: C) Privacy and security

Explanation: In healthcare applications, protecting patient data is paramount, making privacy and security the most critical responsible AI principle to prioritize in this scenario.

Question 4

Which of the following is an example of a document processing AI workload?

  • A) Translating text from English to Spanish
  • B) Extracting information from invoices
  • C) Identifying objects in images
  • D) Converting speech to text

Answer: B) Extracting information from invoices

Explanation: Document processing workloads involve extracting structured information from documents like forms, receipts, invoices, and ID documents.

Question 5

Which responsible AI principle ensures that AI systems are designed to work for people of all abilities and backgrounds?

  • A) Inclusiveness
  • B) Fairness
  • C) Transparency
  • D) Accountability

Answer: A) Inclusiveness

Explanation: Inclusiveness ensures that AI systems are designed to work for everyone, considering diverse user needs and abilities.

Domain 2: Describe Fundamental Principles of Machine Learning on Azure

Question 6

What type of machine learning technique would you use to predict continuous numerical values like house prices?

  • A) Classification
  • B) Clustering
  • C) Regression
  • D) Reinforcement learning

Answer: C) Regression

Explanation: Regression is used to predict continuous numerical values. Classification is used for categorizing data into classes, clustering is used for grouping similar data points, and reinforcement learning is used for decision-making in dynamic environments.

Question 7

In a machine learning dataset, what are the input variables used for prediction called?

  • A) Labels
  • B) Features
  • C) Targets
  • D) Outcomes

Answer: B) Features

Explanation: Features are the input variables used for prediction in machine learning. Labels (also called targets or outcomes) are the output variables being predicted.

Question 8

Which Azure service allows you to automatically select algorithms and tune hyperparameters for your machine learning models?

  • A) Azure Machine Learning Studio
  • B) Azure Databricks
  • C) Automated Machine Learning
  • D) Azure Synapse Analytics

Answer: C) Automated Machine Learning

Explanation: Automated Machine Learning (AutoML) in Azure automatically selects algorithms, performs feature engineering, and tunes hyperparameters to find the best model for your data.

Question 9

What is the purpose of splitting data into training and validation datasets in machine learning?

  • A) To increase the amount of data available for training
  • B) To test the model on data it hasn’t seen during training
  • C) To reduce the computational resources required
  • D) To simplify the model architecture

Answer: B) To test the model on data it hasn’t seen during training

Explanation: Splitting data into training and validation datasets allows you to evaluate how well the model performs on data it hasn’t seen during training, helping to detect overfitting.

Question 10

Which of the following is a characteristic of deep learning compared to traditional machine learning?

  • A) Requires less data
  • B) Always performs better
  • C) Involves multiple layers of neural networks
  • D) Requires less computational resources

Answer: C) Involves multiple layers of neural networks

Explanation: Deep learning involves neural networks with multiple layers (hence “deep”), allowing them to learn complex patterns. It typically requires more data and computational resources than traditional machine learning.

Domain 3: Describe Features of Computer Vision Workloads on Azure

Question 11

Which computer vision task involves identifying and locating objects within images?

  • A) Image classification
  • B) Object detection
  • C) Semantic segmentation
  • D) Optical character recognition

Answer: B) Object detection

Explanation: Object detection involves both identifying what objects are in an image and locating them with bounding boxes. Image classification only categorizes the entire image, semantic segmentation classifies each pixel, and OCR extracts text from images.

Question 12

What Azure service would you use to create a custom image classification model without writing code?

  • A) Azure Machine Learning
  • B) Azure AI Vision
  • C) Azure Custom Vision
  • D) Azure Cognitive Services

Answer: C) Azure Custom Vision

Explanation: Azure Custom Vision allows you to create custom image classification and object detection models through a simple interface without writing code.

Question 13

Which of the following is NOT a capability of the Azure AI Face service?

  • A) Face detection
  • B) Face verification
  • C) Face recognition
  • D) Face animation

Answer: D) Face animation

Explanation: Azure AI Face service provides face detection, verification, and recognition, but does not provide face animation capabilities.

Question 14

What computer vision task would you use to extract text from images?

  • A) Image classification
  • B) Object detection
  • C) Optical character recognition (OCR)
  • D) Facial recognition

Answer: C) Optical character recognition (OCR)

Explanation: OCR is specifically designed to extract text from images, converting images of text into machine-readable text.

Question 15

Which evaluation metric is commonly used to measure the performance of object detection models?

  • A) Accuracy
  • B) Mean Average Precision (mAP)
  • C) Root Mean Square Error (RMSE)
  • D) Perplexity

Answer: B) Mean Average Precision (mAP)

Explanation: Mean Average Precision (mAP) is commonly used to evaluate object detection models, measuring both the precision of the detections and the accuracy of the bounding boxes.

Domain 4: Describe Features of Natural Language Processing (NLP) Workloads on Azure

Question 16

Which NLP task determines the emotional tone of text as positive, negative, or neutral?

  • A) Key phrase extraction
  • B) Named entity recognition
  • C) Sentiment analysis
  • D) Language detection

Answer: C) Sentiment analysis

Explanation: Sentiment analysis specifically focuses on determining the emotional tone or sentiment expressed in text.

Question 17

What Azure service would you use to convert speech to text and text to speech?

  • A) Azure AI Language
  • B) Azure AI Speech
  • C) Azure AI Vision
  • D) Azure Bot Service

Answer: B) Azure AI Speech

Explanation: Azure AI Speech service provides capabilities for speech-to-text, text-to-speech, speech translation, and speaker recognition.

Question 18

Which NLP task identifies entities like people, places, and organizations in text?

  • A) Key phrase extraction
  • B) Named entity recognition
  • C) Sentiment analysis
  • D) Language detection

Answer: B) Named entity recognition

Explanation: Named entity recognition (NER) specifically identifies and categorizes entities like people, places, organizations, dates, etc. in text.

Question 19

What markup language is used to control Speech Synthesis output in Azure AI Speech service?

  • A) HTML
  • B) XML
  • C) SSML
  • D) JSON

Answer: C) SSML

Explanation: Speech Synthesis Markup Language (SSML) is used to control aspects of speech synthesis like pronunciation, volume, pitch, and rate in Azure AI Speech service.

Question 20

Which Azure service would you use to create a knowledge base for a question-answering system?

  • A) Azure AI Language Question Answering
  • B) Azure Bot Service
  • C) Azure AI Speech
  • D) Azure Machine Learning

Answer: A) Azure AI Language Question Answering

Explanation: Azure AI Language Question Answering (formerly QnA Maker) is specifically designed to create knowledge bases for question-answering systems.

Domain 5: Describe Features of Generative AI Workloads on Azure

Question 21

Which of the following is a generative AI model that can create images from text descriptions?

  • A) GPT-4
  • B) DALL-E
  • C) BERT
  • D) Whisper

Answer: B) DALL-E

Explanation: DALL-E is specifically designed to generate images from text descriptions. GPT-4 is primarily a text generation model, BERT is a language understanding model, and Whisper is a speech recognition model.

Question 22

What is a key responsible AI consideration specific to generative AI?

  • A) Hardware efficiency
  • B) Content moderation
  • C) Database integration
  • D) Network latency

Answer: B) Content moderation

Explanation: Content moderation is a key responsible AI consideration for generative AI to prevent the generation of harmful, inappropriate, or misleading content.

Question 23

Which Azure service provides access to large language models like GPT-4?

  • A) Azure Machine Learning
  • B) Azure AI Language
  • C) Azure OpenAI Service
  • D) Azure Bot Service

Answer: C) Azure OpenAI Service

Explanation: Azure OpenAI Service provides access to large language models like GPT-4, GPT-3.5, and other OpenAI models.

Question 24

What is the primary architecture used in modern large language models like GPT?

  • A) Convolutional Neural Networks
  • B) Recurrent Neural Networks
  • C) Transformer
  • D) Generative Adversarial Networks

Answer: C) Transformer

Explanation: The Transformer architecture, with its attention mechanism, is the primary architecture used in modern large language models like GPT.

Question 25

Which of the following is NOT a common application of generative AI?

  • A) Content creation
  • B) Conversational AI
  • C) Network security
  • D) Code generation

Answer: C) Network security

Explanation: While generative AI can be used in security contexts, network security is not one of the primary applications of generative AI. Content creation, conversational AI, and code generation are common applications.

Bonus Questions

Question 26

What factor contributes to the longer training time required for deep learning algorithms compared to traditional machine learning algorithms?

  • A) The ability to use small amounts of data for predictions
  • B) The creation of high-level features from data
  • C) The division of the learning process into smaller steps
  • D) The inherent involvement of numerous layers in deep learning algorithms

Answer: D) The inherent involvement of numerous layers in deep learning algorithms

Explanation: Deep learning models have multiple layers that process data sequentially, which requires more computational resources and time compared to simpler machine learning algorithms.

Question 27

Which Azure AI Document Intelligence prebuilt model is designed to analyze and extract details from legal documents?

  • A) Invoice
  • B) Receipt
  • C) Identity
  • D) Contract

Answer: D) Contract

Explanation: The Contract prebuilt model in Azure AI Document Intelligence is specifically designed to extract information from legal documents and agreements.

Question 28

Which Azure AI Content Safety product would be most suitable for automatically detecting and flagging potentially offensive language in text?

  • A) Analyze Image API
  • B) Jailbreak Risk Detection
  • C) Analyze Text API
  • D) Protected Material Text Detection

Answer: C) Analyze Text API

Explanation: The Analyze Text API in Azure AI Content Safety is designed to detect and flag potentially offensive or inappropriate content in text.

Question 29

What is the best description of machine translation achieved through deep learning?

  • A) Identifying and classifying distinct entities within text
  • B) Speech or text translation from one language to another
  • C) Locating and categorizing objects within pictures
  • D) Examining large volumes of text data to identify patterns

Answer: B) Speech or text translation from one language to another

Explanation: Machine translation specifically refers to the automated conversion of text or speech from one language to another.

Question 30

Which of the following is a clustering machine learning algorithm?

  • A) Linear Regression
  • B) K-means
  • C) Logistic Regression
  • D) Decision Trees

Answer: B) K-means

Explanation: K-means is a clustering algorithm that groups similar data points together. The other options are regression or classification algorithms.