View Full Microsoft AI-901 Exam Dumps and Practice Test Dumps

 

Question 61. Which Azure service provides access to generative AI models for building AI applications?

1. Azure AI Foundry
2. Azure DNS
3. Azure Storage
4. Azure Monitor

Correct Answer: 1. Azure AI Foundry

Explanation:

Azure AI Foundry provides capabilities for developing generative AI and other AI applications using supported models, tools, and Azure services. Developers can use the platform to work with models, evaluate application behavior, integrate data sources, and build AI-powered solutions. It provides an environment that helps organize different components of an AI application rather than requiring developers to manage every capability independently. Azure DNS is designed for name resolution, Azure Storage provides data storage, and Azure Monitor provides monitoring capabilities. Therefore, Azure AI Foundry is the appropriate Azure platform when the goal is to develop applications using generative AI models and related AI capabilities.

Question 62. Which generative AI technique provides instructions without supplying example inputs and outputs?

1. Few-shot prompting
2. Zero-shot prompting
3. Fine-tuning
4. Clustering

Correct Answer: 2. Zero-shot prompting

Explanation:

Zero-shot prompting provides a generative AI model with instructions for completing a task without giving it specific examples of the desired input and output format. The model relies on its existing capabilities and the instructions contained in the prompt to determine how to respond. For example, an application might ask a language model to classify a review as positive or negative without including any sample classifications in the prompt. Few-shot prompting includes examples, while fine-tuning involves adapting a model through additional training. Clustering is an unsupervised machine learning technique. Therefore, providing instructions without examples is known as zero-shot prompting.

Question 63. Which technique supplies a small number of examples in a prompt to guide a model?

1. Zero-shot prompting
2. Few-shot prompting
3. Regression
4. Reinforcement learning

Correct Answer: 2. Few-shot prompting

Explanation:

Few-shot prompting provides a generative AI model with a limited number of examples that demonstrate how a task should be performed. The examples can show the expected relationship between inputs and outputs, desired formatting, classification categories, or response patterns. This can help guide the model when a simple instruction alone does not sufficiently communicate the required behavior. Zero-shot prompting does not provide examples, while regression and reinforcement learning are machine learning approaches rather than prompt techniques. Therefore, when a prompt contains a small number of examples to guide the model’s response, the technique is called few-shot prompting.

Question 64. What is a system message primarily used for in a generative AI application?

1. To define high-level behavior and instructions for the AI assistant
2. To store binary files
3. To configure a DNS server
4. To create a virtual network

Correct Answer: 1. To define high-level behavior and instructions for the AI assistant

Explanation:

A system message is commonly used to establish high-level instructions, behavioral guidance, or context for a generative AI assistant. It can define aspects such as the assistant’s role, response requirements, boundaries, or general behavior before user-specific prompts are processed. This allows developers to provide consistent instructions across interactions. A system message is not intended to function as a binary file store, DNS configuration mechanism, or virtual network definition. Those responsibilities belong to other technologies and services. Therefore, in a generative AI application, the primary purpose of a system message is to establish high-level behavior and instructions for the AI model or assistant.

Question 65. Which AI capability can translate text from one supported language into another?

1. Computer vision
2. Machine learning compute
3. Machine translation
4. Object detection

Correct Answer: 3. Machine translation

Explanation:

Machine translation is an AI capability that automatically converts text from one language into another supported language. It can be used in applications such as multilingual websites, document processing, customer support, communication tools, and localization workflows. The translation system analyzes the source language and generates corresponding content in the target language while attempting to preserve the meaning and context. Computer vision analyzes images, object detection identifies objects within visual content, and machine learning compute provides processing resources rather than performing translation itself. Therefore, when the requirement is to automatically translate text between languages, machine translation is the appropriate AI capability.

Question 66. Which Azure AI service is designed to provide machine translation capabilities?

1. Azure AI Translator
2. Azure AI Vision
3. Azure AI Speech only
4. Azure DNS

Correct Answer: 1. Azure AI Translator

Explanation:

Azure AI Translator provides machine translation capabilities that applications can use to translate supported text between languages. It is useful for scenarios such as multilingual applications, websites, customer communications, document workflows, and other solutions that need automated language conversion. Translation can be integrated into applications through supported APIs and Azure services. Azure AI Vision focuses on visual analysis, Azure AI Speech provides speech-related capabilities, and Azure DNS handles domain-name resolution. Therefore, when an application needs a dedicated Azure service for automatically translating text between supported languages, Azure AI Translator is the appropriate choice.

Question 67. Which Azure AI service provides capabilities for extracting information from forms and structured documents?

1. Azure AI Document Intelligence
2. Azure AI Speech
3. Azure AI Search
4. Azure DNS

Correct Answer: 1. Azure AI Document Intelligence

Explanation:

Azure AI Document Intelligence provides document analysis capabilities that can extract structured information from documents such as forms, invoices, receipts, and other supported document types. It can identify fields, values, tables, and other relevant content depending on the model or feature being used. This capability is useful for automating document-processing workflows that would otherwise require manual data entry. Azure AI Speech is designed for speech processing, Azure AI Search provides information retrieval, and Azure DNS provides domain-name resolution. Therefore, Azure AI Document Intelligence is the appropriate service when an organization needs to extract structured information from forms and documents.

Question 68. Which computer vision task identifies text contained within an image?

1. Optical character recognition
2. Clustering
3. Regression
4. Sentiment analysis

Correct Answer: 1. Optical character recognition

Explanation:

Optical character recognition, commonly called OCR, identifies and extracts text from images or scanned documents. OCR can convert visual representations of printed or handwritten characters into machine-readable text, allowing applications to process information that originally existed only in an image. Common scenarios include scanning receipts, extracting text from forms, reading signs, and processing scanned documents. Clustering groups similar data points, regression predicts numerical values, and sentiment analysis evaluates opinions or emotional tone in text. Therefore, when an application needs to recognize written or printed text contained within an image, optical character recognition is the appropriate computer vision capability.

Question 69. Which AI workload can identify whether a face appears in an image?

1. Speech synthesis
2. Face detection
3. Text summarization
4. Regression

Correct Answer: 2. Face detection

Explanation:

Face detection is a computer vision capability used to identify the presence and location of human faces in an image. A detection system can determine where faces appear within visual content and may provide location information such as bounding regions. Face detection should be distinguished from face recognition or identification, which involves determining whether a detected face corresponds to a particular known individual and introduces additional privacy and responsible AI considerations. Speech synthesis works with audio generation, text summarization processes written content, and regression predicts numerical values. Therefore, identifying whether and where a human face appears in an image is a face detection workload.

Question 70. Which Azure AI service is designed for analyzing and understanding spoken language?

1. Azure AI Speech
2. Azure AI Search
3. Azure AI Vision
4. Azure Storage

Correct Answer: 1. Azure AI Speech

Explanation:

Azure AI Speech provides capabilities for working with spoken language and audio. Depending on the feature being used, it can support speech-to-text, text-to-speech, speech translation, and other speech-related scenarios. Applications can use these capabilities to create voice interfaces, transcription systems, accessibility solutions, and conversational experiences. Azure AI Search is focused on search and retrieval, Azure AI Vision analyzes visual content, and Azure Storage provides data storage. Therefore, when the requirement is to process spoken language or convert between speech and text, Azure AI Speech is the appropriate Azure AI service.

Question 71. Which machine learning method learns from rewards and penalties received while interacting with an environment?

1. Supervised learning
2. Unsupervised learning
3. Reinforcement learning
4. Regression

Correct Answer: 3. Reinforcement learning

Explanation:

Reinforcement learning is a machine learning approach in which an agent interacts with an environment and learns through feedback in the form of rewards or penalties. The objective is generally to learn a strategy or policy that maximizes the cumulative reward over time. This approach can be applied to certain decision-making and control problems where actions influence subsequent states and outcomes. Supervised learning uses labeled examples, unsupervised learning discovers patterns in unlabeled data, and regression predicts numerical values. Therefore, a learning approach based on interactions with an environment and reward-based feedback is reinforcement learning.

Question 72. What is the main purpose of a test dataset in machine learning?

1. To train model parameters
2. To provide an independent evaluation of the final model
3. To define DNS records
4. To store application secrets

Correct Answer: 2. To provide an independent evaluation of the final model

Explanation:

A test dataset is used to evaluate the performance of a machine learning model after the development and training process has been completed. Ideally, test data should not be used to train or repeatedly tune the model because doing so can make the evaluation less representative of performance on genuinely unseen data. The test results can provide an estimate of how the final model may perform when it encounters new examples. Training data is used to learn model parameters, while validation data is commonly used during development and tuning. Therefore, the primary purpose of a test dataset is to provide an independent evaluation of the final model.

Question 73. Which issue occurs when a model performs well on training data but poorly on unseen data?

1. Overfitting
2. Underfitting
3. Data encryption
4. Load balancing

Correct Answer: 1. Overfitting

Explanation:

Overfitting occurs when a machine learning model learns the training data too specifically and fails to generalize effectively to new, unseen examples. Such a model may achieve very high training performance while producing significantly worse results on validation or test data. Overfitting can be caused by excessive model complexity, insufficient training data, noisy data, or other factors. Underfitting represents a different problem in which the model is too simple to capture important patterns and performs poorly even on training data. Data encryption and load balancing are infrastructure concepts unrelated to this machine learning behavior. Therefore, strong training performance combined with weak unseen-data performance indicates overfitting.

Question 74. Which issue occurs when a model is too simple to learn important patterns in the training data?

1. Overfitting
2. Underfitting
3. Clustering
4. Classification

Correct Answer: 2. Underfitting

Explanation:

Underfitting occurs when a machine learning model is too simple or insufficiently trained to capture important relationships in the data. As a result, the model may perform poorly on both the training dataset and new data. This differs from overfitting, where the model may perform very well on training data but poorly on unseen examples because it has learned training-specific patterns too closely. Developers can address underfitting through approaches such as using a more suitable model, adding informative features, reducing excessive regularization, or improving training. Therefore, when a model is too simple to learn important patterns from the training data, the issue is underfitting.

Question 75. Which Azure service can monitor the performance and health of Azure resources and applications?

1. Azure Monitor
2. Azure Translator
3. Azure AI Vision
4. Azure Blob Storage

Correct Answer: 1. Azure Monitor

Explanation:

Azure Monitor provides monitoring and observability capabilities for Azure resources, applications, and supported connected environments. It can collect and analyze metrics, logs, and other monitoring data and can support alerting when configured conditions occur. Monitoring is important for understanding application performance, detecting operational problems, and responding to resource or service conditions. Azure Translator provides language translation, Azure AI Vision analyzes images, and Azure Blob Storage stores unstructured data. Therefore, when an organization needs to monitor the performance and health of Azure resources and applications, Azure Monitor is the appropriate service.

Question 76. Which Azure capability can help identify inappropriate or harmful content generated by an AI application?

1. Azure AI Content Safety
2. Azure DNS
3. Azure Load Balancer
4. Azure Blob Storage

Correct Answer: 1. Azure AI Content Safety

Explanation:

Azure AI Content Safety provides capabilities for detecting and managing potentially harmful content in AI applications. It can help organizations evaluate user inputs and model-generated outputs against supported harmful-content categories and implement appropriate moderation or filtering workflows. This can be particularly important for applications that interact directly with users or generate content automatically. Azure DNS provides name resolution, Azure Load Balancer distributes network traffic, and Azure Blob Storage provides object storage. Therefore, when the requirement is to identify potentially inappropriate or harmful content in an AI solution, Azure AI Content Safety is the relevant Azure capability.

Question 77. Which practice can help reduce the risk of exposing sensitive information in an AI application’s prompts?

1. Data minimization
2. Increasing screen resolution
3. DNS round robin
4. Network compression

Correct Answer: 1. Data minimization

Explanation:

Data minimization involves providing and retaining only the information that is necessary for a particular purpose. In an AI application, limiting sensitive information included in prompts can reduce the amount of personal or confidential data exposed to the model and related processing components. Organizations should also apply appropriate access controls, security measures, retention policies, and data-handling practices based on the application’s requirements. Increasing screen resolution, DNS round robin, and network compression do not directly address prompt-level exposure of sensitive information. Therefore, data minimization is an important practice for reducing unnecessary exposure of sensitive information in AI prompts.

Question 78. Which Azure AI capability can evaluate an AI application’s responses against predefined quality or safety criteria?

1. Evaluation
2. DNS forwarding
3. Storage replication
4. DHCP reservation

Correct Answer: 1. Evaluation

Explanation:

AI evaluation involves systematically assessing an AI application’s behavior and outputs against defined criteria. Depending on the evaluation scenario, criteria can include response quality, relevance, groundedness, safety, or other application-specific measures. Evaluation helps developers identify weaknesses and compare different prompts, models, or application configurations before and after changes. It is an important part of developing reliable AI applications because a system should not be judged only by whether it produces an output. DNS forwarding, storage replication, and DHCP reservations are infrastructure functions unrelated to evaluating AI responses. Therefore, evaluation is the appropriate capability for assessing AI application outputs against defined criteria.

Question 79. Which concept describes connecting a generative AI model to external information so responses can use that information?

1. Grounding
2. Overfitting
3. Clustering
4. Classification

Correct Answer: 1. Grounding

Explanation:

Grounding connects a generative AI model with relevant external information so that generated responses can incorporate context from trusted data sources. For example, a business application can retrieve information from internal documents, databases, or search indexes and provide that information to a language model as context for generating a response. This can help the application answer questions using organization-specific or more current information rather than relying solely on what the model learned during training. Overfitting is a machine learning issue, clustering groups similar data, and classification predicts categories. Therefore, connecting external information to improve contextual AI responses is known as grounding.

Question 80. Which Azure service can provide a managed environment for deploying an AI model as an online endpoint?

1. Azure Machine Learning
2. Azure DNS
3. Azure Storage
4. Azure Files

Correct Answer: 1. Azure Machine Learning

Explanation:

Azure Machine Learning provides capabilities for deploying machine learning models to managed online endpoints that applications can call for real-time inference. This allows a trained model to be made available through a service interface so applications can submit input and receive predictions. The platform also supports other parts of the machine learning lifecycle, including training, model management, evaluation, and monitoring capabilities. Azure DNS provides name resolution, Azure Storage stores data and artifacts, and Azure Files provides managed file shares. Therefore, when an organization needs a managed environment for deploying a machine learning model as an online inference endpoint, Azure Machine Learning is the appropriate service.