AI-900 Practice Exam 56 — 20 Questions

Question 1

A content creator uses Azure OpenAI to generate unique story ideas for a fantasy novel. They want the output to be highly creative and unpredictable, avoiding common clichés. Which parameter should they primarily increase to achieve this?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Temperature. Increasing the Temperature parameter makes the model's output more random and less deterministic, which is ideal for generating highly creative and unpredictable story ideas. A higher temperature (e.g., 0.9–1.0) increases the probability of sampling less likely tokens, reducing repetition and clichés.

Answer

Top p

Answer

Frequency penalty

Answer

Presence penalty

Question 2

A data scientist trains a decision tree model to predict customer churn. The model achieves 99% accuracy on the training data but only 80% on the test data. Which concept best explains this performance difference?

Accepted Answer

Overfitting. The model's high accuracy on training data (99%) but significantly lower accuracy on test data (80%) indicates that it has memorized the training data rather than learning generalizable patterns. This is the classic symptom of overfitting, where the decision tree captures noise and outliers in the training set, leading to poor performance on unseen data.

Answer

Underfitting

Answer

Bias-variance tradeoff

Answer

Cross-validation

Question 3

What is the difference between Azure AI Speech and Azure AI Language?

Accepted Answer

Azure AI Speech handles audio processing; Azure AI Language processes and understands text. Azure AI Speech is designed to process audio input, converting speech to text (speech-to-text) and text to speech (text-to-speech), as well as enabling speaker recognition and real-time translation. Azure AI Language, on the other hand, processes and understands text by providing capabilities such as sentiment analysis, key phrase extraction, language detection, and question answering. Option B correctly captures this fundamental division: Speech handles audio processing, while Language handles text understanding.

Answer

Azure AI Speech is for transcription only; Azure AI Language handles all other NLP tasks

Answer

Azure AI Speech works only in English; Azure AI Language supports multiple languages

Answer

They are the same service with different pricing tiers

Question 4

A bank is developing an AI system to automatically approve or reject small business loan applications. The bank wants to ensure that the system does not unfairly discriminate against applicants based on their age, gender, or ethnicity. Which Microsoft responsible AI principle should most directly guide the design and evaluation of this system?

Accepted Answer

Fairness. The bank's goal is to prevent discrimination based on age, gender, or ethnicity in loan approvals. The Fairness principle directly addresses this by requiring AI systems to treat all groups equitably and to mitigate biases in training data and model predictions. This principle guides the design and evaluation of the system to ensure that outcomes are not skewed by protected attributes.

Answer

Reliability and safety

Answer

Privacy and security

Answer

Inclusiveness

Question 5

A medical research team needs to analyze thousands of clinical trial reports to extract specific medical terms like disease names, symptoms, and medications. They want to use an Azure AI Language feature that is pre-trained on medical data and requires no custom training. Which feature should they use?

Accepted Answer

Text Analytics for Health (Healthcare NLP). Text Analytics for Health (Healthcare NLP) is a pre-trained Azure AI Language feature specifically designed to extract medical entities such as disease names, symptoms, medications, and treatment details from unstructured clinical text. It requires no custom training and is built on medical ontologies like UMLS, making it ideal for analyzing thousands of clinical trial reports without additional model development.

Answer

Key Phrase Extraction

Answer

Named Entity Recognition (NER)

Answer

Sentiment Analysis

Question 6

What is the Azure AI Speech service's 'speaker diarization' feature?

Accepted Answer

Identifying and labeling different speakers in a multi-speaker audio recording. Speaker diarization is the process of partitioning an audio stream into homogeneous segments according to the speaker identity. It answers the question 'who spoke when?' by assigning a unique label (e.g., Speaker 1, Speaker 2) to each segment in a multi-speaker recording. This is a built-in capability of Azure AI Speech, often used in meeting transcription and call center analytics.

Answer

Converting a speaker's voice to a different language in real time

Answer

Measuring the speaking speed (words per minute) of each speaker

Answer

Filtering out background speakers from a primary speaker's recording

Question 7

Which Azure service provides a no-code/low-code drag-and-drop interface for building machine learning pipelines?

Accepted Answer

Azure Machine Learning Designer. Azure Machine Learning Designer is the correct answer because it provides a drag-and-drop, no-code/low-code visual interface for building, testing, and deploying machine learning pipelines. Users can connect pre-built modules for data transformation, model training, and scoring without writing code, making it ideal for rapid prototyping and operationalization of ML workflows.

Answer

Azure AI Custom Vision

Answer

Azure AI Language Studio

Answer

Azure Databricks

Question 8

A retail company uses overhead cameras to monitor shelf inventory in a store. They want to build a system that automatically detects whether a shelf section is empty or stocked, and specifically identify product categories (e.g., 'soft drinks', 'chips', 'canned goods') and count the number of items in each category. The company has a large set of labeled images showing different shelf states. Which Azure Computer Vision service should they use to build this custom detection and counting solution?

Accepted Answer

Custom Vision object detection. Custom Vision object detection is the correct choice because it allows the company to train a model on their labeled images to detect and localize specific product categories (e.g., 'soft drinks', 'chips') and count items within each category. Unlike pre-built Computer Vision features, Custom Vision enables custom object detection with bounding boxes and classification, which directly supports the requirement for detecting shelf states and counting items per category.

Answer

Computer Vision Image Analysis with dense captioning

Answer

Optical Character Recognition (OCR)

Answer

Azure Machine Learning with a pre-trained YOLO model

Question 9

An insurance company uses an AI system to automatically process and approve or reject claims. The system sometimes rejects valid claims because the uploaded documents are in slightly different formats (e.g., PDF vs. scanned images). The company wants to minimize these errors. Which Microsoft responsible AI principle is most directly relevant to addressing this issue?

Accepted Answer

Reliability and safety. The issue is that the AI system fails to process valid claims due to variations in document formats (PDF vs. scanned images), which is a reliability and safety problem. The system should be robust enough to handle input variations and consistently produce correct outcomes. Microsoft's Reliability and safety principle focuses on ensuring AI systems operate reliably, safely, and consistently under expected conditions, directly addressing the need to minimize such errors.

Answer

Fairness

Answer

Inclusiveness

Answer

Transparency

Question 10

A data scientist is training a binary classification model to detect rare equipment failures from sensor data. The dataset contains 99.5% normal operation readings and only 0.5% failure readings. The model currently predicts all readings as 'normal' and achieves 99.5% accuracy on the test set. The business requires the model to identify at least 80% of actual failures. Which data-level technique should the data scientist use to most directly address the class imbalance?

Accepted Answer

Oversample the minority class (failure examples). Oversampling the minority class (failure examples) directly addresses the severe class imbalance by creating synthetic copies or duplicates of the rare failure instances. This balances the training dataset, allowing the model to learn patterns associated with failures rather than always predicting the majority class. With a balanced dataset, the model can be trained to meet the business requirement of identifying at least 80% of actual failures, even though overall accuracy may decrease.

Answer

Undersample the majority class (normal examples)

Answer

Use precision as the optimization metric

Answer

Reduce the complexity of the model

Question 11

A data scientist trains a regression model on a dataset with 100 features and 10,000 samples. The model achieves a low training error but a much higher error on a held-out test set. Which approach is most likely to improve the model's generalization performance?

Accepted Answer

Reduce the number of features or apply regularization. The model exhibits high variance (overfitting), as indicated by low training error but high test error. Reducing the number of features or applying regularization (e.g., L1/L2) directly constrains model complexity, forcing it to learn more general patterns rather than memorizing noise. This is the standard approach to improve generalization in regression models.

Answer

Increase the complexity of the model by adding more layers or parameters

Answer

Add more training data

Answer

Use a different train-test split ratio like 80-20 instead of 70-30

Question 12

What is the 'presence penalty' parameter in Azure OpenAI API calls?

Accepted Answer

A flat penalty discouraging repetition of any token already present in the response. The 'presence penalty' parameter in Azure OpenAI API calls applies a flat penalty to any token that has already appeared in the response so far, reducing the model's likelihood of repeating that token. This helps generate more diverse and less repetitive text by discouraging the reuse of tokens already present in the output sequence.

Answer

A parameter requiring AI systems to acknowledge their presence as AI to users

Answer

A parameter indicating whether the AI is present online or offline

Answer

The minimum number of characters that must be present in a response

Question 13

A data scientist trains a classification model on a dataset of 10,000 labeled emails to distinguish spam from non-spam. The model achieves 99% accuracy on the training data but only 70% accuracy on a held-out test set. Which term best describes this situation?

Accepted Answer

B) Overfitting. The model performs exceptionally well on training data (99% accuracy) but poorly on unseen test data (70% accuracy), which is the classic symptom of overfitting. Overfitting occurs when the model learns noise and specific patterns in the training set rather than generalizing to new data, often due to excessive model complexity or insufficient regularization.

Answer

A) Underfitting

Answer

C) Bias-variance tradeoff

Answer

D) Regularization

Question 14

A retail company wants to predict which customers are likely to cancel their subscription in the next 30 days. What ML task type is this?

Accepted Answer

Binary classification to predict whether each customer will cancel or stay. This is a binary classification task because the goal is to predict one of two mutually exclusive outcomes for each customer: either they will cancel (churn) or stay (not churn) within the next 30 days. Binary classification algorithms, such as logistic regression or decision trees, are specifically designed to assign each input to one of two discrete labels based on learned patterns from historical data.

Answer

Clustering to identify similar customer segments

Answer

Regression to predict the customer's lifetime value

Answer

Generative AI to write personalized retention emails

Question 15

What is 'Azure AI Vision's Read API' and what makes it superior for OCR?

Accepted Answer

An advanced OCR service handling multi-page PDFs, handwriting, and complex layouts with word-level coordinates. Azure AI Vision's Read API is an advanced OCR service that extracts text from images and documents, including multi-page PDFs, handwritten text, and complex layouts. It is superior because it returns word-level bounding box coordinates and confidence scores, enabling precise text localization and structured output for downstream processing.

Answer

The standard API for reading data from Azure Storage accounts and databases

Answer

An API for reading audio content and converting it to text transcripts

Answer

A feature for reading the metadata of image files stored in Azure Blob Storage

Question 16

What is the difference between 'training' and 'inference' in machine learning?

Accepted Answer

Training creates models from data; inference uses trained models to make predictions. Training is the phase where a machine learning model learns patterns from labeled or unlabeled data by adjusting its internal parameters (e.g., weights in a neural network) to minimize a loss function. Inference is the subsequent phase where the trained model applies those learned patterns to new, unseen data to generate predictions or classifications. In Azure Machine Learning, training typically involves running a script on a compute target (e.g., a GPU cluster) and registering the resulting model, while inference is performed by deploying that model as a real-time endpoint or batch pipeline.

Answer

Training is for testing models; inference is for training them

Answer

They are the same process with different names for clarity

Answer

Training is for image models; inference is for text models

Question 17

A hotel booking website wants to automatically analyze guest-submitted photos of hotel rooms to verify if they contain common amenities such as a bed, a desk, and a chair. They want to use a prebuilt Azure AI service without any custom training. Which feature should they use?

Accepted Answer

Object Detection. Object Detection (prebuilt) is the correct choice because it can identify and locate multiple specific objects (bed, desk, chair) within an image by drawing bounding boxes around them. This prebuilt Azure AI Vision feature requires no custom training and directly supports detecting common amenities in hotel room photos.

Answer

Optical Character Recognition (OCR)

Answer

Image Analysis (prebuilt)

Answer

Handwriting OCR

Question 18

A customer service team wants to analyze thousands of call transcripts to identify common complaints and understand whether customer sentiment is positive, negative, or neutral. They plan to use prebuilt Azure AI Language features without any custom training. Which combination of features should they use?

Accepted Answer

Key phrase extraction and sentiment analysis. The customer service team needs to identify common complaints (which requires extracting key phrases from the call transcripts) and understand sentiment polarity (positive, negative, or neutral). Azure AI Language's prebuilt key phrase extraction returns the main talking points and recurring terms, while sentiment analysis assigns a sentiment score and labels per sentence or document. Both features are available out-of-the-box without any custom training, directly meeting the stated requirements.

Answer

Entity recognition and text translation

Answer

Language detection and summarization

Answer

PII detection and conversation analysis

Question 19

A marketing team uses Azure OpenAI Service to generate marketing copy. They notice the generated text is often repetitive, using the same phrases and words multiple times. Which parameter should they increase to directly reduce this repetition?

Accepted Answer

Frequency penalty. Frequency penalty directly reduces repetition by penalizing tokens that have already appeared in the generated text. A higher frequency penalty value (e.g., 0.5 to 1.0) decreases the likelihood of the model reusing the same phrases or words, making the output more diverse and less repetitive.

Answer

Temperature

Answer

Top-p

Answer

Max tokens

Question 20

What is 'language detection' in Azure AI Language?

Accepted Answer

Identifying the natural human language of input text with a confidence score. Language detection in Azure AI Language identifies the natural human language of input text and returns a confidence score indicating the likelihood of the detected language being correct. This is a core capability of the Language service, using pre-trained machine learning models to classify text into over 100 languages without requiring any training data from the user.

Answer

Automatically identifying which programming language a code snippet is written in

Answer

Checking whether text contains any language that violates community guidelines

Answer

Translating text from any language into the user's preferred language