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

 

Question 161. Which Azure service provides a centralized environment for developing generative AI applications?

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

Correct Answer: 2. Azure AI Foundry

Explanation:

Azure AI Foundry provides an environment for building, testing, evaluating, and deploying generative AI applications. It brings together models, prompts, evaluation capabilities, and application development resources in a unified workspace. Developers can use available foundation models and connect AI capabilities to application data and services. The platform is particularly useful when an organization needs to move beyond simply calling a model and wants to manage the broader development lifecycle. Azure Storage and Azure DNS provide infrastructure-related capabilities, while Azure Firewall focuses on network security. Therefore, Azure AI Foundry is the appropriate choice when the requirement is a centralized environment for developing generative AI solutions.

Question 162. What is the main purpose of a system message in a generative AI application?

1. To store training data
2. To create database indexes
3. To define behavior and instructions for the model
4. To monitor network traffic

Correct Answer: 3. To define behavior and instructions for the model

Explanation:

A system message provides high-level instructions that guide how a generative AI model should behave during an interaction. It can establish the assistant’s role, communication style, rules, limitations, or task-specific behavior. For example, a system message could instruct a model to act as a customer-support assistant and answer only questions related to company products. This is different from a user message, which normally contains the specific request from the person interacting with the application. System instructions therefore help developers control the model’s behavior consistently across multiple interactions. They are not intended for storing datasets, creating indexes, or monitoring network traffic.

Question 163. Which technique provides a generative AI model with a few examples of the desired task?

1. Few-shot prompting
2. Data normalization
3. Clustering
4. Feature scaling

Correct Answer: 1. Few-shot prompting

Explanation:

Few-shot prompting involves giving a generative AI model a small number of examples that demonstrate how a task should be completed. The examples can show an input and the expected output, allowing the model to infer the pattern and apply it to a new input. For instance, a prompt might provide several examples of customer comments classified as positive or negative before asking the model to classify another comment. This technique can improve consistency without requiring the model itself to be retrained. Clustering, normalization, and feature scaling are different data-processing or machine-learning concepts and do not describe the process of supplying examples directly within a prompt.

Question 164. What is the primary benefit of grounding a generative AI response with external data?

1. It eliminates the need for prompts
2. It increases computer memory
3. It replaces the AI model
4. It helps provide responses based on relevant information

Correct Answer: 4. It helps provide responses based on relevant information

Explanation:

Grounding connects a generative AI model with relevant external information so that its responses can be based on specific data rather than relying only on information contained in the model’s training. For example, a company chatbot can retrieve current product documentation before generating an answer about a product feature. This can improve relevance and help reduce unsupported responses. Grounding does not eliminate prompts or replace the underlying model. It also does not directly increase computer memory. In many AI applications, retrieval systems such as search indexes are used to locate useful information, which is then supplied to the model as context for generating the response.

Question 165. Which search approach is particularly useful for finding documents based on the meaning of a query?

1. Exact filename matching
2. Semantic search
3. File compression
4. Port scanning

Correct Answer: 2. Semantic search

Explanation:

Semantic search focuses on the meaning and intent of a query rather than relying only on exact keyword matches. This makes it useful when users phrase questions differently from the wording contained in the indexed documents. For example, a user might search for “ways to protect account information,” while a document discusses “customer data security.” Semantic search can recognize the relationship between these concepts and return relevant information. Azure AI Search supports semantic capabilities that can improve the relevance of search results. Exact matching, compression, and port scanning serve entirely different purposes and do not provide meaning-based document retrieval.

Question 166. What do embeddings represent in many AI search solutions?

1. Network addresses
2. Database passwords
3. Numerical representations of data meaning
4. File permissions

Correct Answer: 3. Numerical representations of data meaning

Explanation:

Embeddings are numerical representations that capture meaningful characteristics and relationships within data such as text, images, or other content. In AI-powered search, text can be converted into vectors, allowing a system to compare the semantic similarity between a user’s query and stored documents. Content with related meanings can have vectors that are closer together in the embedding space even when the exact words differ. This makes embeddings useful for vector search and retrieval-augmented generation scenarios. Embeddings are not network addresses, passwords, or permission records. Their primary role is to represent information in a mathematical form that AI systems can compare and process.

Question 167. Which Azure service can be used to search indexed organizational content for an AI application?

1. Azure AI Search
2. Azure Virtual Network
3. Azure DNS
4. Azure Key Vault

Correct Answer: 1. Azure AI Search

Explanation:

Azure AI Search is designed to index and retrieve information from organizational data sources. It can support keyword, semantic, and vector-based search scenarios, making it useful for AI applications that need to retrieve relevant information before generating an answer. For example, an organization can index internal documents and allow an AI assistant to retrieve relevant passages when answering employee questions. This retrieval capability is especially important in retrieval-augmented generation applications. Azure Virtual Network provides networking capabilities, Azure DNS handles name resolution, and Azure Key Vault is designed for managing secrets and keys. Therefore, Azure AI Search is the service most directly associated with searching indexed organizational content.

Question 168. Which scenario is an example of retrieval-augmented generation (RAG)?

1. Training a neural network from scratch
2. Compressing an image
3. Creating a firewall rule
4. Retrieving company documents and providing them to a language model for an answer

Correct Answer: 4. Retrieving company documents and providing them to a language model for an answer

Explanation:

Retrieval-augmented generation combines information retrieval with generative AI. In a typical RAG workflow, a user’s question is first used to retrieve relevant information from a search index or knowledge source. The retrieved content is then provided to a language model as additional context. The model uses that context to generate a response. This approach is useful when an application needs to answer questions about private, specialized, or frequently changing information without requiring the foundation model to be retrained whenever the source data changes. Training a neural network, compressing images, and configuring firewall rules are separate activities and do not describe the RAG process.

Question 169. Which Azure AI capability can help detect potentially harmful content in user-generated text or images?

1. Azure AI Content Safety
2. Azure DNS
3. Azure Data Box
4. Azure Load Balancer

Correct Answer: 1. Azure AI Content Safety

Explanation:

Azure AI Content Safety provides capabilities for identifying and managing potentially harmful content in applications that process text and images. It can help organizations detect categories of unsafe material and apply appropriate controls based on their application requirements. This is especially important when users can submit content to an AI-powered application or when an application generates responses that need to be evaluated. Content Safety is part of responsible AI practices because it can help developers establish safeguards around AI interactions. Azure DNS, Azure Data Box, and Azure Load Balancer address networking, data transfer, and traffic distribution needs respectively, rather than content moderation.

Question 170. Which computer vision task assigns a category to an entire image?

1. Object detection
2. Image classification
3. Speech recognition
4. Entity recognition

Correct Answer: 2. Image classification

Explanation:

Image classification assigns one or more categories or labels to an entire image. For example, a model might classify an image as containing a dog, cat, vehicle, or landscape. The important characteristic is that classification determines what the image represents without necessarily identifying the precise location of individual objects. Object detection goes further by identifying objects and locating them with bounding boxes. Speech recognition processes spoken language, while entity recognition identifies meaningful entities in text. Understanding the distinction between image classification and object detection is important when selecting an AI vision technique for a particular application requirement.

Question 171. What does OCR primarily extract from an image or scanned document?

1. Network packets
2. Audio signals
3. Written or printed text
4. Database relationships

Correct Answer: 3. Written or printed text

Explanation:

Optical Character Recognition, or OCR, is used to detect and extract text from images and scanned documents. It can recognize printed or handwritten characters depending on the capabilities of the selected service and input. OCR is useful for digitizing paper documents, extracting text from photographs, and making scanned information searchable. For example, an organization could process scanned forms and extract the visible text into structured application data. OCR does not analyze network packets, process audio signals, or identify database relationships. In Azure AI solutions, OCR capabilities can be used as part of document and image-processing workflows.

Question 172. Which Azure AI Language capability identifies names of people, organizations, or locations in text?

1. Named entity recognition
2. Image segmentation
3. Speech synthesis
4. Object detection

Correct Answer: 1. Named entity recognition

Explanation:

Named entity recognition, or NER, identifies entities within text and categorizes them according to types such as person, organization, location, date, or other supported categories. For example, in the sentence “Microsoft opened an office in London,” an NER system could identify Microsoft as an organization and London as a location. This capability is useful for extracting structured information from unstructured text. Image segmentation and object detection are computer vision tasks, while speech synthesis converts text into spoken audio. NER therefore provides the functionality required when an application needs to locate and categorize meaningful entities within written language.

Question 173. Which Azure AI Speech capability converts spoken audio into text?

1. Text-to-speech
2. Speech-to-text
3. Machine translation
4. Image analysis

Correct Answer: 2. Speech-to-text

Explanation:

Speech-to-text converts spoken language in an audio stream or recording into written text. It is commonly used for transcription, voice-controlled applications, meeting notes, call-center processing, and accessibility scenarios. The service analyzes the audio input and produces text representing the recognized speech. Text-to-speech performs the opposite operation by converting written text into synthesized audio. Machine translation focuses on converting content between languages, while image analysis works with visual information. Therefore, when an application needs to capture what a person says and represent it as written text, the appropriate Azure AI Speech capability is speech-to-text.

Question 174. What is the purpose of text-to-speech in an AI application?

1. To convert spoken audio into text
2. To classify images
3. To convert text into synthesized speech
4. To identify database keys

Correct Answer: 3. To convert text into synthesized speech

Explanation:

Text-to-speech converts written text into computer-generated spoken audio. This capability can make applications more accessible and useful in situations where users prefer to listen rather than read. Examples include virtual assistants, navigation systems, educational applications, and automated announcements. A text-to-speech service can generate audio using a selected voice and language, depending on the supported capabilities. Speech-to-text is the reverse process, converting spoken language into written text. Image classification deals with visual data, while database keys are related to data storage. Therefore, text-to-speech is the appropriate technology when an application needs to speak generated or stored textual content.

Question 175. Which Azure service is designed to translate text between supported languages?

1. Azure AI Translator
2. Azure Monitor
3. Azure Blob Storage
4. Azure Virtual Machines

Correct Answer: 1. Azure AI Translator

Explanation:

Azure AI Translator provides machine translation capabilities for converting text from one supported language to another. It can be incorporated into applications that need multilingual communication, such as websites, customer-support tools, document workflows, and messaging systems. Translation services can process source text and return an equivalent version in the requested target language. Azure Monitor is used for monitoring applications and resources, Azure Blob Storage provides object storage, and Azure Virtual Machines provide compute resources. Therefore, when the primary requirement is automated language translation, Azure AI Translator is the service designed specifically for that purpose.

Question 176. Which Azure service can extract structured fields from invoices and receipts?

1. Azure DNS
2. Azure AI Document Intelligence
3. Azure Firewall
4. Azure Virtual Network

Correct Answer: 2. Azure AI Document Intelligence

Explanation:

Azure AI Document Intelligence is designed to analyze documents and extract information from forms, invoices, receipts, identity documents, and other supported document types. Instead of treating a document as only an image containing text, document-processing models can identify fields and values and return structured information that applications can use. For example, an invoice-processing application may extract invoice numbers, dates, vendor information, and totals. This reduces the need for manual data entry. Azure DNS, Azure Firewall, and Azure Virtual Network provide infrastructure and networking capabilities, so they are not appropriate for extracting structured fields from business documents.

Question 177. What is the main purpose of an Azure Machine Learning pipeline?

1. To automate a sequence of machine-learning steps
2. To replace all storage services
3. To configure DNS records
4. To create firewall policies

Correct Answer: 1. To automate a sequence of machine-learning steps

Explanation:

An Azure Machine Learning pipeline organizes and automates a sequence of machine-learning activities. A pipeline can include steps such as preparing data, training a model, evaluating results, and producing outputs. Automating these steps helps make machine-learning workflows more repeatable and easier to manage. For example, an organization could create a pipeline that retrieves training data, performs preprocessing, trains a model, and evaluates its performance. Pipelines do not replace storage services or perform networking tasks such as DNS configuration or firewall management. Their purpose is to coordinate machine-learning workflow activities in a structured and repeatable manner.

Question 178. When should an online endpoint generally be used for machine-learning inference?

1. When predictions are needed interactively with low latency
2. When no predictions are required
3. Only when storing archived documents
4. Only for configuring networks

Correct Answer: 1. When predictions are needed interactively with low latency

Explanation:

An online endpoint is appropriate when an application needs predictions or model responses interactively, often with low latency. For example, a customer-facing application might send an input to a deployed machine-learning model and immediately receive a prediction. Online inference is useful when individual requests arrive continuously and responses are expected quickly. Batch inference is different because it is designed to process larger quantities of data without requiring an immediate response for each individual request. Storage, networking, and archival tasks are separate concerns. Selecting the correct endpoint type depends largely on how frequently predictions are requested and how quickly the application needs the results.

Question 179. Which responsible AI principle focuses on protecting personal information?

1. Fairness
2. Transparency
3. Privacy and security
4. Accountability

Correct Answer: 3. Privacy and security

Explanation:

Privacy and security focus on protecting personal information and ensuring that data used by an AI system is handled appropriately. AI applications may process names, contact details, financial information, documents, or other sensitive business data, so appropriate access controls and data-handling practices are important. Techniques such as limiting unnecessary data collection, controlling permissions, and protecting stored information can reduce privacy risks. Fairness focuses on avoiding unjustified differences in outcomes, transparency concerns understanding and communicating how systems work, and accountability concerns assigning responsibility for AI systems and their outcomes. Therefore, privacy and security directly address the protection of personal information.

Question 180. What should be investigated when a deployed model’s performance decreases because the input data has changed over time?

1. The monitor’s screen resolution
2. Model or data drift
3. The DNS domain name only
4. The keyboard layout

Correct Answer: 2. Model or data drift

Explanation:

When the characteristics of incoming data change over time, the performance of a machine-learning model can decrease. This situation can involve data drift, where the distribution or characteristics of input data change, or model-related performance changes caused by evolving real-world conditions. Monitoring deployed models helps organizations identify these changes and determine whether investigation, retraining, or updating is necessary. For example, a model trained using historical customer behavior may become less accurate if customer behavior changes substantially. Screen resolution, keyboard layout, and DNS naming do not explain changes in model prediction quality. Therefore, drift should be investigated when input patterns evolve and model performance declines.