Summary of "AI 900 Exam Q&A #1 - Azure AI Fundamentals"

Video overview / logistics

Presenter: Bibek
Content: 250 questions and answers for AI-900 (Azure AI Fundamentals)
Exam info:

Passing score = 700 / 1000. Fee varies by country (presenter mentions £69 in the UK).
Download / purchase: Bibek offers the full question set on his “Buy Me a Coffee” page; members get 50% off.

Main topics covered (questions 1–10)

Purpose and scope of conversational AI
Basic dataset handling for model training and evaluation
Microsoft guiding principles for responsible AI (accountability; reliability & safety; privacy & security)
Choosing the correct Azure resource for a given AI task (Form Recognizer for receipts)
NLP tasks for monitoring news (entity recognition and sentiment analysis)
Distinguishing classification use cases from regression/clustering tasks
Azure OpenAI models and primary capabilities (DALL·E for image generation)
Basic ML workflow step ordering (load data first)
Responsible AI considerations for handling unusual/missing values (reliability & safety)
Possible channels a bot can connect to (Teams, email, web chat, Alexa)

Detailed Q&A (questions 1–10)

Detecting a change in hospital infection rates is a scenario of conversational AI.
- Correct answer: False
You are testing a machine learning model — how should you split data for training and evaluation?
- Options:
  - A: features for training / labels for evaluation
  - B: randomly split rows into training and remaining rows for evaluation
  - C: labels for training / features for evaluation
  - D: split columns
- Correct answer: B — randomly split the data (rows) into training and evaluation sets
Drag-and-drop: Microsoft guiding principles for responsible AI — match descriptions to principles.
- “Implementing processes so that decisions made by the AI system can be overridden by humans” → Accountability
- “Ensure AI systems operate as designed, respond to unanticipated conditions and resist harmful manipulations” → Reliability and safety
- “Provide consumers with information and controls over the collection, use, and storage of their data” → Privacy and security
Team member plans to use the Form Recognizer pre-built receipt model — which Azure resource should you recommend?
- Options: Face service; Form Recognizer; Computer Vision
- Correct answer: Form Recognizer
System to scan news for articles about customers, alert on negative articles, add positive articles to a press book — which NLP tasks to use?
- Correct answers: Entity recognition (to find customer mentions) and Sentiment analysis (to classify articles as negative/positive)
Select the correct use case for classification.
- Options included:
  - Predict number of cups of coffee (regression)
  - Predict minutes to run a race (regression)
  - Predict whether someone uses a bicycle to travel to work based on distance (classification)
  - Group images by color (clustering)
- Correct answer: Predicting whether someone uses a bicycle to travel to work based on distance from home to work (classification)
Which Azure OpenAI model can generate original images from natural language?
- Options: DALL·E; Whisper; Embeddings; GPT-3.5; GPT-4
- Correct answer: DALL·E
You have a dataset and need to build an Azure ML classification model to identify defective products — what should you do first?
- Options: create a clustering model; split data into training/testing; create a classification model; load the dataset
- Correct answer: Load the dataset
Developing an AI system that will be ingesting unusual and missing values — which Microsoft responsible AI principle applies?
- Options: Inclusiveness; Reliability and safety; Fairness
- Correct answer: Reliability and safety
What channels can a bot be connected to? (select all correct) - Correct answers: Microsoft Teams; Email; Web chat; Alexa