Microsoft Azure AI Fundamentals Practice Test (AI-900)

Generative models can produce new content such as marketing slogans by learning from existing data. They can generate creative and brand-aligned slogans that resonate with the target audience, making this approach appropriate.

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.

The Azure AI Language service is designed for analyzing text using natural language processing techniques, including key phrase extraction, sentiment analysis, and entity recognition. It allows users to process large volumes of text data to gain insights into the content.

The Azure AI Speech service focuses on speech-to-text and text-to-speech capabilities and is not optimized for text analysis tasks.

Azure Cognitive Search is used for adding search functionalities to applications but doesn't provide advanced text analytics out of the box.

Azure Machine Learning service is a platform for building, training, and deploying machine learning models, which would require custom development effort to perform these specific text analysis tasks.

The Azure AI Face Detection service is purpose-built to locate faces in an image and optionally return detailed facial attributes such as head pose, facial landmarks, occlusion, and quality-for-recognition scores. These specialized outputs make it the best fit for applications that need fine-grained facial analysis.

The Azure AI Vision service can detect faces but is optimized for general image analysis and does not provide the same depth of facial-attribute data. Azure Video Indexer targets video scenarios, and Azure AI Text Analytics works exclusively with textual content, so neither meets the stated requirement.

Image Classification is when a model assigns a single label to an entire image based on its overall content. It does not identify the presence or location of multiple objects within the image.

Object Detection on the other hand involves identifying and locating multiple objects within an image.

Optical Character Recognition (OCR) extracts text from images.

Semantic Segmentation assigns a class label to each pixel in the image, effectively delineating object boundaries.

Classification - This is the correct answer. Fraud detection is a classification task because the model needs to classify each transaction as either fraudulent or legitimate, which involves assigning data points to predefined categories or labels.

Regression is used for predicting continuous numerical values for example sales forecasts or prices, not for classifying transactions into categories such as "fraudulent" or "legitimate."

Clustering is an unsupervised learning technique used to group data based on similarities, but it is not suitable for determining whether a transaction is fraudulent or legitimate, which requires labeled data and a classification approach.

Dimensionality Reduction is used to reduce the number of features in the data, typically for improving performance or visualization, but it is not a task in itself for determining fraud or legitimacy.

They create new content by modeling the structure of the training data - This is the correct answer. Generative AI models create new and original content by learning the underlying structure, patterns, and relationships within the training data. This allows them to generate creative outputs, like story ideas, that are inspired by but not directly copied from the original material.

They store exact copies of data for direct replication - This is characteristic of retrieval-based systems, which retrieve exact content from a database or dataset. This does not apply to generative AI, which creates novel content rather than copying it.

They categorize input data into predefined classes - This is the function of classification models, which categorize data into specific classes based on predefined labels. It does not involve generating new content.

They detect anomalies and irregularities in data - This describes anomaly detection models, which focus on identifying outliers or unusual patterns in data, not generating new and original content.

Entity Recognition is used to identify and extract specific entities like dates, monetary values and company names from text. It transforms unstructured text into structured data by categorizing elements into predefined entity types.

Key Phrase Extraction highlights important terms and phrases but doesn't categorize them into entities.

Sentiment Analysis gauges the emotional tone of the text.

Language Modeling predicts word sequences for tasks like text generation.

Analyzing images and extracting visual information - This is the correct answer. The Azure AI Vision service provides developers with capabilities to analyze images and extract visual information. This includes tasks such as image classification, object detection, and optical character recognition (OCR), helping developers to gain insights from images.

Transcribing spoken language into text - This is the functionality of speech recognition services, not the Azure AI Vision service. Speech recognition is focused on converting spoken language into written text.

Translating text between different languages - This is the functionality of Azure Translator service, which is designed for translating text between languages, not related to image analysis.

Performing sentiment analysis on text data - This is the functionality of text analytics services, which are used for tasks like sentiment analysis, key phrase extraction, and language detection, but it is not related to image analysis.

Azure Automated Machine Learning is designed to automate the process of selecting the most appropriate algorithms and tuning hyperparameters for your dataset and problem type. It iteratively trains models with different algorithms and parameters to find the best performing model.

Azure Machine Learning Designer provides a drag-and-drop interface for building models but requires you to select algorithms and parameters manually.

Azure Machine Learning Studio Notebooks offer a coding environment for custom model development, which may not save time for those with limited expertise.

Azure Cognitive Services provide pre-built AI services but are not used for custom model training.

The Codex family of models in Azure OpenAI Service was specifically trained on a massive corpus of public code and natural language to translate descriptions into executable code. While newer models like GPT-4 also have strong coding capabilities, the term 'Codex' refers to the models originally specialized for this purpose, as reflected in the AI-900 exam objectives.

Using a training dataset that covers all relevant demographic groups exposes the model to varied examples and reduces the chance that it learns patterns that favor or exclude particular populations.

Reducing the sampling temperature only makes outputs more deterministic; it does not address underlying bias in the learned parameters.

Removing outliers that correspond to minority groups shrinks representation and tends to worsen bias.

Simply increasing the depth or parameter count of the network can even amplify existing bias if the data remain unbalanced.

Knowledge Mining - This is the correct answer. Knowledge mining is the most appropriate AI workload for indexing and analyzing a vast collection of unstructured documents, such as text files, PDFs, and images. It involves extracting insights, identifying patterns, and making the information easily searchable to uncover hidden insights and help users find relevant data quickly.

Natural Language Processing (NLP) is useful for processing and understanding text, knowledge mining includes broader capabilities for extracting insights from various unstructured data types (e.g., text, PDFs and images), making it more suitable for this scenario.

Document Intelligence is focused on understanding and processing documents, such as extracting data from structured documents like forms or invoices, but it may not cover the full breadth of knowledge mining, which includes a wider range of unstructured data sources.

Content personalization is aimed at tailoring content or recommendations to individual users based on their preferences or behaviors. It does not directly address the need to index and analyze a large collection of unstructured documents.

The ability to recognize and extract printed and handwritten text from images is a key feature of optical character recognition (OCR) solutions. OCR allows for the conversion of text within images into editable and searchable data formats.

The other options describe features of different computer vision solutions, classifying images into categories is related to image classification, detecting faces and analyzing features pertains to facial detection and analysis, and segmenting and identifying objects within images is a feature of object detection solutions.

The Azure AI Vision service provides a wide range of computer vision functionalities, including image analysis, object detection, facial detection and analysis, and optical character recognition (OCR). This makes it suitable for detecting customer demographics (using face analysis), recognizing products on shelves (object detection), and reading text from signs (OCR).

The Azure AI Face Detection service specializes in facial detection and recognition but does not handle object detection or OCR.

The Azure AI Anomaly Detector service is designed for identifying anomalies in time-series data, not for image or video analysis.

The Azure AI Language service focuses on natural language processing tasks and does not provide image analysis capabilities.

The design decisions focus on Inclusiveness. Microsoft's Inclusiveness principle states that AI systems should "empower everyone and engage all people, regardless of their backgrounds" and be inclusive "for people of all abilities." Adding screen-reader friendly labels, keyboard navigation, and descriptive alt-text removes barriers for users with disabilities, directly addressing this principle. The other options target different Responsible AI areas: Fairness concerns bias and equal treatment, Transparency is about making the system understandable, and Accountability relates to human oversight.

Automated model selection and hyperparameter tuning - This is the correct answer. Azure's automated machine learning feature includes automated model selection and hyperparameter tuning, which automatically selects the best algorithm for your predictive model and optimizes hyperparameters. This streamlines the model development process and helps improve the model's performance.

Automated feature engineering to create new features from existing data - While automated feature engineering is helpful for transforming raw data into meaningful features, it does not directly involve selecting the best algorithm or tuning hyperparameters, which are the focus of the question.

Deployment of models as scalable web services - Deployment helps you make your model available for real-time or batch predictions, but it does not focus on the algorithm selection or hyperparameter tuning during the model development phase.

Visualization of data using interactive dashboards - Visualization is a key part of understanding data and model performance, but it does not directly contribute to selecting algorithms or tuning hyperparameters for the predictive model.

Deep Learning - This is the correct answer. Deep learning is the most appropriate technique for analyzing unstructured data like images and speech, as it can automatically learn hierarchical representations from raw data. Deep learning models, especially neural networks, excel at processing and extracting meaningful features from complex data without requiring extensive manual feature engineering.

Clustering Algorithms are used for grouping data into clusters based on similarity, but they are not designed to automatically learn hierarchical representations from raw unstructured data like images and speech.

Regression Algorithms are used for predicting continuous numerical outcomes based on input features, but they are not suited for learning hierarchical representations from unstructured data.

Decision Trees are used for classification or regression tasks by splitting data into decision rules based on features. They require predefined features and are less effective than deep learning for automatically learning hierarchical representations from raw, unstructured data like images and speech.

Codex is the Azure OpenAI model specifically designed for code generation tasks, allowing developers to transform natural language prompts into code in various programming languages.

GPT-3's text-davinci-003 is powerful for natural language understanding and generation but is not optimized for code generation.

DALL·E is used for image generation.

Azure's Computer Vision API is intended for analyzing visual content, not generating code.

Generative AI models are designed to produce new content-text, images, or other data-by learning patterns in their training data. Predictive models focus on forecasting future outcomes, classification models assign inputs to predefined categories, and reinforcement-learning systems improve through iterative feedback. Therefore, the ability to synthesize original content is what uniquely distinguishes generative AI.