Microsoft Azure AI Fundamentals Practice Test (AI-900)

The Image Tagging feature of the Azure AI Vision service is the correct choice. It is designed to analyze images and return content tags for thousands of recognizable objects, living beings, scenery, and actions. This allows for the automatic generation of descriptive keywords for cataloging and searching a collection of images. Optical Character Recognition (OCR) is used to extract text from images, Face Detection is used to locate human faces, and Spatial Analysis is used to analyze the presence and movement of people in video feeds.

Producing new data similar to the training data - This is the correct answer. Generative AI models are designed to generate new data that is similar to the training data, such as creating new images, text, or music. They learn patterns from the input data and use this knowledge to produce new, original content.

Classifying data into predefined categories - This is a characteristic of classification models, not generative AI. Classification models categorize data into specific labels or classes, but they do not generate new data.

Detecting anomalies in data patterns - Anomaly detection is typically done by models designed for identifying outliers or unusual patterns in data, not generative AI models.

**Compressing data into lower-dimensional representations **- This is the purpose of dimensionality reduction techniques (e.g., PCA), which aim to reduce the number of features in data while preserving important information. It is not related to generative AI.

Features: Age, Education Level, Years of Experience; Label: Salary - This is the correct answer. In this scenario, the features are the input variables used to predict the target, which is the 'Salary'. So, 'Age', 'Education Level', and 'Years of Experience' are the features, and 'Salary' is the label (the variable you're trying to predict).

Features: Salary; Label: Age, Education Level, Years of Experience - This is incorrect because 'Salary' is the target variable, not a feature used to predict other variables.

Features: Age, Salary, Education Level, Years of Experience; Label: Salary - This is incorrect because 'Salary' is the target variable, and it should not be included as a feature when you're predicting it.

Features: Age, Education Level; Label: Salary, Years of Experience - This is incorrect because 'Salary' is the target variable, and you should not treat multiple columns (like both 'Salary' and 'Years of Experience') as labels.

The DALL·E model in Azure OpenAI Service is specifically designed for generating images from text descriptions. It allows developers to create realistic images and art from natural language prompts.

The GPT-3 model is designed for text generation tasks and cannot produce images.

Azure Cognitive Services Computer Vision API is intended for analyzing images, not generating them.

**Azure Machine Learning **would allow you to build a custom image generation model but it would require significant effort and resources.

Generating synthetic data to augment datasets - This is the correct answer. Generative AI is commonly used to create synthetic data, which can help augment existing datasets.

Image recognition in security systems - This is a task for computer vision, not generative AI.

Categorizing customer feedback into topics - This task is handled by classification or clustering models, not generative AI.

Predicting stock prices using historical data - This is a predictive modeling task, not generative AI.

Sentiment Analysis evaluates text to determine the emotional tone, categorizing it as positive, negative or neutral. This helps the company determine customer satisfaction directly from reviews.

Key Phrase Extraction identifies important terms but doesn't assess emotions.

Named entity recognition extracts specific entities like names or locations.

Language Detection determines the language used in the text.

Azure OpenAI exposes GPT models for natural-language generation through the Completions endpoint family (including chat/completions for conversational use). This endpoint accepts a text prompt and returns a text completion, making it the correct choice for tasks such as drafting email replies. The Embeddings endpoint only returns vector representations of text, the Image generation endpoint produces images, and the Audio transcription endpoint converts speech to text-none of these create the required textual response.

The Azure AI Vision service provides extensive image analysis capabilities, including identifying objects within images and extracting text through optical character recognition (OCR).

Azure AI Face service is specialized for detecting and analyzing human faces, but doesn't support general object identification or text extraction.

Azure AI Document Intelligence is designed to extract structured data from forms and documents but is less suitable for general image analysis tasks.

Azure Speech to Text service is used for converting spoken language into text and does not process images.

This scenario represents a Generative AI workload because the system is generating new original artwork that imitates the style of existing artists. Generative AI focuses on producing new data that shares characteristics with the training data.

Computer Vision workloads involve interpreting and analyzing visual information but not generating new images.

Knowledge Mining workloads are about extracting insights from existing data, not creating new content.

Content Moderation workloads deal with identifying and filtering inappropriate content. Therefore, the most suitable workload in this case is Generative AI.

A model specialized in code generation - This is the correct answer. For generating code snippets from natural language descriptions, the developer should use a model like Codex, which is specifically designed for code generation. It can understand natural language inputs and generate corresponding code in various programming languages.

A model specialized in image generation - This type of model, like DALL·E, is designed for generating images from text descriptions, not for generating code.

A model specialized in sentiment analysis - Sentiment analysis models are used to assess the emotional tone of text, not to generate code based on natural language descriptions.

A model specialized in generating long-form text - While models like GPT are capable of generating text, they are not specifically optimized for code generation, making them less ideal for this task compared to a model like Codex.

Azure Translator is designed specifically for translating text between languages while maintaining the original context and meaning. It utilizes advanced neural machine translation to handle idiomatic expressions and nuances in language.

The Azure Text Analytics service's key phrase extraction identifies important terms in text but does not translate text.

The Azure Speech service for speech recognition transcribes spoken words into text but does not translate text.

Azure Text Analytics for language detection identifies the language of a given text but does not perform translation.

Natural Language Processing (NLP) is used to analyze and understand human language in text or speech form. Since the company wants to analyze textual customer reviews to determine sentiment, NLP techniques are appropriate for this task. Computer Vision focuses on visual data like images and videos, which doesn't apply to text reviews. Predictive Maintenance and Time Series Forecasting involve predicting equipment failures and future values based on time-series data, respectively, neither of which relate to analyzing text reviews for sentiment.

Provide detailed documentation on how the model makes predictions - This is the correct answer. To ensure stakeholders understand how the AI model makes decisions, providing clear and detailed documentation on how the model works, including its features, decision-making process, and limitations, is essential for transparency and trust.

Use a complex ensemble model to maximize predictive accuracy - While ensemble models can improve predictive accuracy, they do not directly address the concern about stakeholders understanding how the model makes decisions.

Encrypt the model’s parameters to protect intellectual property - Encrypting the model’s parameters may be important for intellectual property protection, but it does not help in making the decision-making process more transparent or understandable to stakeholders.

Limit access to the model to senior management - Limiting access to the model only to senior management reduces transparency and makes it harder for other stakeholders to understand how decisions are made.

Regression is the appropriate technique for predicting continuous numerical values, such as the number of customer service calls. It models the relationship between dependent and independent variables to forecast future values.

Classification is used for predicting categorical outcomes

Clustering groups data points based on similarity without prior labels.

Entity recognition is used to identify and classify key elements in text into predefined categories such as names of people, locations and organizations. This makes it suitable for extracting such specific information from large text datasets.

Key Phrase Extraction identifies important phrases but does not categorize them into entities.

Sentiment Analysis determines the emotional tone of text.

Text Summarization condenses text to key points but doesn't extract specific entities.

Entity recognition is the appropriate NLP technique for this task. It involves identifying and classifying key elements in text into predefined categories such as names of persons, organizations, locations, dates and monetary values. This allows the company to extract structured data from unstructured text.

Key Phrase Extraction identifies important terms and phrases but doesn't categorize them into specific entity types.

Sentiment Analysis determines the emotional tone behind a body of text.

Translation converts text from one language to another.

The Azure AI Face service is the specialized service for detecting human faces and analyzing detailed facial features, including the location of facial landmarks and head pose.

The Azure AI Vision service can perform general image analysis, including basic face detection, but it does not provide the same level of detailed facial attribute analysis as the dedicated Face service.

The Azure AI Speech service is used for speech processing tasks, not image analysis.

The Azure AI Video Indexer service focuses on extracting insights from video content, not detailed analysis of still images.

Computer Vision focuses on processing and interpreting visual information from images or videos. It is the most suitable choice for analyzing video surveillance to detect suspicious activities.

Natural Language Processing deals with understanding and generating human language, which is not applicable to visual data.

Knowledge Mining involves extracting insights from large volumes of structured and unstructured data, typically text-based.

Generative AI is concerned with creating new content rather than analyzing existing footage.

They should use Computer Vision. This workload enables machines to interpret and understand visual information from images or videos. In this scenario, the retail company aims to analyze images from cameras to identify and categorize products, which is a typical application of Computer Vision.

Natural Language Processing deals with understanding and processing human language, which is not applicable here.

Knowledge Mining involves extracting information from large volumes of text but doesn't directly address image analysis.

Generative AI focuses on creating new content, such as text or images, and is not suited for product identification tasks.

Ensuring that decisions are explainable and can be traced back to responsible parties promotes accountability. This allows organizations to understand how decisions are made, take responsibility for outcomes, address any issues, and comply with regulations. Practices like prioritizing performance over transparency, relying solely on automation without oversight, or keeping algorithms confidential can hinder accountability because they obscure the decision-making process and who is responsible.