Microsoft Azure AI Fundamentals AI-900 (AI-900) — Questions 1–75

Why this answer

How to eliminate wrong answers

Option A is wrong because Image Analysis does not return file size or dimensions; those are basic metadata properties handled by storage services, not the computer vision API. Option C is wrong because the service returns multiple category labels, tags, and descriptions, not just a single category label for the entire image. Option D is wrong because Azure AI Vision does not generate 3D point clouds; that capability is associated with depth-sensing cameras or specialized 3D reconstruction services, not the 2D image analysis API.

Why this answer

How to eliminate wrong answers

Option A is wrong because fairness focuses on mitigating bias and ensuring equitable treatment across demographic groups, not on documenting human-AI decision interactions. Option B is wrong because reliability and safety concern the system's consistent performance and robustness against failures, not the legal documentation of overrides. Option C is wrong because transparency involves explaining how the AI system works and its limitations, but the core need here is to document who made the final decision and why, which falls under accountability.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes batch transcription or translation of pre-recorded audio, which is a different Azure service (Batch Transcription API) and not real-time. Option C is wrong because generating subtitles for pre-existing videos is a batch or offline task, often handled by Azure Video Indexer or Media Services, not the real-time Speech Translation API. Option D is wrong because it describes text-to-speech conversion in the same language, which is a separate capability (Text-to-Speech API) and does not involve translation between languages.

Why this answer

How to eliminate wrong answers

Option A is wrong because Key Phrase Extraction identifies main topics or concepts in text, not personal identifiers like names or SSNs. Option B is wrong because Sentiment Analysis determines the emotional tone (positive, negative, neutral) of text, not sensitive data. Option D is wrong because Language Detection identifies the language of the text (e.g., English, Spanish), not PII.

Why this answer

How to eliminate wrong answers

Option A is wrong because the principle of Fairness focuses on avoiding bias and ensuring equitable treatment across demographic groups, not on providing multiple interaction modalities. Option B is wrong because the principle of Reliability and Safety concerns system performance, accuracy, and harm prevention, not accessibility features. Option D is wrong because the principle of Privacy and Security deals with data protection and user consent, not with interface accessibility or inclusivity.

Why this answer

How to eliminate wrong answers

Option A is wrong because reliability and safety focus on ensuring the AI system performs consistently and without harm, not on providing interpretable explanations for individual decisions. Option B is wrong because fairness addresses bias and equitable treatment across patient groups, not the ability to understand why a specific recommendation was made. Option D is wrong because accountability refers to assigning responsibility for the AI system's outcomes and governance, not the technical interpretability of its decisions.

Why this answer

How to eliminate wrong answers

Option A is wrong because temperature controls randomness in token selection, not repetition; higher temperature increases creativity but does not prevent phrase repetition. Option C is wrong because presence penalty penalizes tokens based on whether they have appeared at all, not how often, so it reduces topic repetition but not multiple occurrences of the same phrase. Option D is wrong because max tokens limits the total length of the output, not the repetition of phrases within it.

Why this answer

How to eliminate wrong answers

Option A is wrong because checking whether a text-to-speech voice pronounces technical terms correctly is a text-to-speech quality evaluation, not a feature of pronunciation assessment, which analyzes human speech input. Option C is wrong because detecting regional accents to route calls is a separate capability often handled by language identification or custom speech models, not by pronunciation assessment, which scores against a single native norm. Option D is wrong because generating a list of commonly mispronounced words is not a direct output of pronunciation assessment; it scores individual utterances but does not aggregate mispronunciation lists across a language.

Why this answer

How to eliminate wrong answers

Option A is wrong because key phrase extraction identifies important words or phrases in text but does not assign a sentiment label. Option B is wrong because language detection identifies the language of the text (e.g., English, Spanish) and has no capability to classify sentiment. Option D is wrong because entity recognition identifies named entities (e.g., people, places, organizations) and does not evaluate the emotional tone of the text.

Why this answer

How to eliminate wrong answers

Option A is wrong because Custom Text Classification requires you to provide labeled training data to build a custom model, which is unnecessary here since a pre-trained healthcare-specific NER model already exists. Option C is wrong because Key Phrase Extraction identifies general important phrases (e.g., 'patient history') but does not understand or categorize medical entities like diagnoses or medications. Option D is wrong because Sentiment Analysis determines the emotional tone (positive, negative, neutral) of text, which is irrelevant to extracting structured medical terms from clinical notes.

Why this answer

How to eliminate wrong answers

Option A is wrong because the Temperature parameter controls the randomness of token selection in the model's output, not content safety. Option B is wrong because the Top-p parameter (nucleus sampling) limits the cumulative probability of token choices to influence output diversity, not filter harmful content. Option D is wrong because the Max-tokens parameter sets a hard limit on the length of the generated response, but has no ability to detect or block prohibited content.

Why this answer

How to eliminate wrong answers

Option B (Image Analysis) is wrong because it focuses on describing visual content (objects, scenes, tags) and does not extract text from images. Option C (Face detection) is wrong because it identifies human faces and facial attributes, not textual data from documents. Option D (Custom Vision) is wrong because it is used to train custom image classification or object detection models, not to perform general-purpose text extraction from handwritten forms.

Why this answer

How to eliminate wrong answers

Option A is wrong because fairness focuses on ensuring AI systems do not discriminate against groups or individuals based on attributes like race or gender, not on explaining how a decision was made. Option B is wrong because reliability and safety concern the system's ability to function consistently and safely under expected conditions, not the right to understand model outputs. Option D is wrong because privacy and security deal with protecting data from unauthorized access and ensuring confidentiality, not with providing explanations of model reasoning.

Why this answer

How to eliminate wrong answers

Option A is wrong because Optical Character Recognition (OCR) is designed to extract text from images, not to detect or locate physical objects like people or pets. Option C is wrong because Image Captioning generates a human-readable description of an image's content but does not provide bounding box coordinates for individual objects. Option D is wrong because Custom Vision requires custom training with labeled images to detect specific objects, which contradicts the requirement to use a prebuilt service without any custom training.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Language focuses on natural language processing tasks like sentiment analysis, key phrase extraction, and language detection, but it does not provide translation capabilities. Option C is wrong because Azure OpenAI Service is a generative AI service for tasks like text generation and summarization, not a dedicated translation service, and using it for translation would be inefficient and less accurate than a purpose-built translator. Option D is wrong because Azure Speech Service handles speech-to-text and text-to-speech conversion, not text translation, and is irrelevant for translating written emails.

Why this answer

How to eliminate wrong answers

Option A is wrong because Fairness addresses bias and equitable treatment across demographic groups, not the technical accuracy or reliability of predictions. Option C is wrong because Privacy and Security focus on protecting patient data from unauthorized access or breaches, not on the correctness of the AI's diagnostic output. Option D is wrong because Inclusiveness ensures the system is usable by diverse populations, but does not directly govern the precision or dependability of the model's inferences.

Why this answer

How to eliminate wrong answers

Option A is wrong because both binary and multi-class classification models can output categorical labels or numeric probabilities; the distinction is not about output type but the number of classes. Option C is wrong because classification tasks are not inherently tied to data modality—binary classification can be applied to text (e.g., sentiment analysis) and multi-class to images (e.g., object recognition). Option D is wrong because accuracy depends on the dataset and problem complexity, not the number of classes; multi-class problems often have lower baseline accuracy due to more classes, but neither type is universally more accurate.

Why this answer

How to eliminate wrong answers

Option A is wrong because a system prompt is not an error message; error messages in Azure OpenAI are returned as HTTP status codes (e.g., 400 for bad request) or specific error objects, not as system prompts. Option C is wrong because the user's first message is a 'user prompt' or 'user input,' not a system prompt; the system prompt is set by the developer before any user interaction. Option D is wrong because there is no command to restart an AI model instance in Azure OpenAI; model instances are stateless and managed via deployment endpoints, and a system prompt does not trigger a restart.

Why this answer

How to eliminate wrong answers

Option A is wrong because smart cropping is not about processing speed or GPU acceleration; it is about content-aware cropping guided by AI. Option C is wrong because smart cropping does not automatically remove faces for privacy; that would be a separate feature like face blurring or anonymization. Option D is wrong because smart cropping does not remove background noise or irrelevant context; it preserves the most important content, which may include background elements if they are salient.

Why this answer

How to eliminate wrong answers

Option A is wrong because Temperature controls the randomness of the entire output distribution by scaling logits uniformly, not the probability of a specific token. Option C is wrong because Top P (Nucleus Sampling) selects from a cumulative probability mass of tokens, but it does not allow targeting or reducing the chance of a single token like 'Corporation'. Option D is wrong because Frequency Penalty reduces the likelihood of tokens that have already appeared in the generated text, not the likelihood of a specific token regardless of context.

Why this answer

How to eliminate wrong answers

Option A is wrong because Key Phrase Extraction identifies single words or short phrases (e.g., 'policy', 'document') but cannot extract full answer sentences or handle question-answering logic. Option B is wrong because Conversational Language Understanding (CLU) is a task-oriented intent/entity extraction service that requires manual definition of intents and utterances, not direct answer extraction from documents. Option D is wrong because Azure OpenAI on your data uses large language models to generate answers, but it is not a dedicated extractive QA service; it relies on a retrieval-augmented generation (RAG) approach that may hallucinate or rephrase answers, whereas Custom Question Answering strictly extracts verbatim spans from the provided documents.

Why this answer

How to eliminate wrong answers

Option A is wrong because sentiment analysis evaluates the emotional tone (positive, negative, neutral) of text, not the specific topics or themes discussed. Option C is wrong because entity recognition identifies named entities such as people, organizations, or locations, not general product features or abstract topics like 'customer support'. Option D is wrong because language detection identifies the language of the text (e.g., English, Spanish), which is irrelevant to extracting topical content from reviews.

Why this answer

How to eliminate wrong answers

Option B is wrong because Fairness addresses bias against demographic groups (e.g., race, gender), not environmental conditions like weather. Option C is wrong because Transparency concerns explainability and disclosure of how the AI works, not its operational robustness in different weather. Option D is wrong because Privacy and security focus on data protection and unauthorized access, not system performance or safety under varied conditions.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Search is not a web crawling service; it indexes your own data (e.g., from Azure Blob Storage, Cosmos DB) using built-in indexers, not publicly available web content. Option C is wrong because Azure AI Search is not used for compliance or resource configuration scanning; that is the role of Azure Policy or Azure Security Center. Option D is wrong because Azure AI Search is a standalone search service with its own indexing and query capabilities, not a plugin that simply adds full-text search to Azure SQL databases.

Why this answer

How to eliminate wrong answers

Option A is wrong because key phrase extraction identifies general key terms and phrases in text, not predefined entity types like dates or monetary amounts. Option C is wrong because sentiment analysis determines the emotional tone (positive, negative, neutral) of text, not the extraction of structured information. Option D is wrong because language detection identifies the language of the text (e.g., English, Spanish), not specific entities within the document.

Why this answer

How to eliminate wrong answers

Option A is wrong because creating a custom voice persona that sounds different from standard Azure voices describes Custom Neural Voice, not Custom Speech. Option C is wrong because configuring speech recognition to only accept voice commands from authorised users relates to speaker recognition or voice authentication, not custom speech recognition. Option D is wrong because building a custom programming language for writing speech processing scripts is not a feature of Azure AI Speech; custom speech focuses on training acoustic and language models, not scripting languages.

Why this answer

How to eliminate wrong answers

Option A is wrong because Inclusiveness focuses on designing AI systems that empower and engage a diverse range of human users, not on correcting biased decision-making outcomes that already exist. Option C is wrong because Transparency is about ensuring systems are understandable and their decisions can be explained, but it does not directly mandate equitable treatment or fix the disparity in priority assignment. Option D is wrong because Accountability holds individuals or organizations responsible for the system's behavior, but it does not prescribe the specific corrective action of eliminating bias in triage decisions.

Why this answer

How to eliminate wrong answers

Option A is wrong because sharing model configuration with other Azure subscriptions is done via resource sharing or RBAC, not through an export operation; export produces a file, not a subscription transfer. Option C is wrong because moving the model to Azure Machine Learning would involve registering the model in AML, but Custom Vision's export feature is specifically for downloading a file for offline use, not for moving to another Azure service. Option D is wrong because submitting the model for Microsoft certification review is not a feature of Custom Vision; certification is unrelated to the export functionality.

Why this answer

How to eliminate wrong answers

Option B (max_tokens) is wrong because it limits the length of the generated text, not the randomness or diversity of the output. Option C (top_p) is wrong because it controls nucleus sampling, which selects from the smallest set of tokens whose cumulative probability exceeds a threshold; while it also affects diversity, it is not the primary parameter for controlling randomness—temperature is the standard choice. Option D (frequency_penalty) is wrong because it reduces the likelihood of repeating the same tokens or phrases, which addresses repetition rather than overall randomness or tonal variation.

Why this answer

How to eliminate wrong answers

Option A is wrong because a hardware accelerator for AI model training refers to specialized chips like GPUs (e.g., NVIDIA A100) or FPGAs, not a software-based AI assistant; copilots run on existing hardware and do not accelerate training. Option C is wrong because a database for storing conversation history is a data store (e.g., Azure Cosmos DB or a vector database), not an AI assistant; copilots may use such databases to maintain context but are not themselves databases. Option D is wrong because a software testing tool for AI models (e.g., Azure Machine Learning's model evaluation or fairness assessment tools) is used to validate model performance, not to assist users in completing tasks via natural language.

Why this answer

How to eliminate wrong answers

Option A is wrong because the pre-built Computer Vision object detection API is trained on generic object classes and cannot reliably detect small, partially visible penguins that blend into the background. Option C is wrong because dense captioning generates descriptive text for image regions, not bounding boxes or counts for specific objects like penguins. Option D is wrong because image classification assigns a single label to the entire image, whereas the team needs to detect and count multiple penguins per image, which requires object detection.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes online learning or Bayesian updating of model parameters with new data, not hyperparameter tuning. Option C is wrong because it describes adaptive learning rate methods like Adam or SGD with momentum, which adjust the learning rate during training based on gradients, not a search over hyperparameter space. Option D is wrong because it describes active learning or uncertainty sampling for labeling, which selects data points for human annotation, not hyperparameter optimization.

Why this answer

How to eliminate wrong answers

Option B (Frequency penalty) is wrong because it reduces repetition by penalizing tokens that have already appeared, but it does not control determinism or creativity; it can still allow creative language. Option C (Presence penalty) is wrong because it encourages the model to talk about new topics by penalizing tokens that have appeared at all, which can actually increase variability and speculative language. Option D (Top_p) is wrong because it controls nucleus sampling (the cumulative probability threshold for token selection) and, while it can reduce randomness, it does not guarantee maximal determinism like setting temperature to 0 does; a low top_p still allows some randomness within the selected nucleus.

Why this answer

How to eliminate wrong answers

Option A is wrong because connecting multiple AI models in a processing pipeline describes a model orchestration or ensemble architecture (e.g., using Azure Logic Apps to chain a translator with a summarizer), not a prompting technique. Option C is wrong because linking multiple conversation turns to maintain context refers to session management or multi-turn dialog state tracking (e.g., using conversation history in a chatbot), not a single-prompt reasoning method. Option D is wrong because training a model using sequential text data only describes a training data format (e.g., sequence-to-sequence learning or autoregressive pretraining), not a prompting strategy applied at inference time.

Why this answer

How to eliminate wrong answers

Option A is wrong because fine-tuning involves updating the model's weights through additional training on a dataset, which is not what is happening here—the examples are provided only in the prompt at inference time. Option C is wrong because training a custom model from scratch requires a massive dataset, significant compute resources, and is not a technique used with Azure OpenAI Service for this task; the developer is using a pre-trained model. Option D is wrong because reinforcement learning from human feedback (RLHF) is a training process used to align model behavior based on human preferences, not a method for providing examples in a single prompt to generate code.

Why this answer

How to eliminate wrong answers

Option A is wrong because a system message is not an error notification; Azure OpenAI uses HTTP status codes (e.g., 503 Service Unavailable) or specific error responses to indicate service unavailability, not a system prompt. Option C is wrong because the model does not send automated confirmation messages; user input is acknowledged implicitly through the model's response, and there is no built-in 'received' confirmation mechanism in the chat completion API. Option D is wrong because the first message a user sends is a 'user message' (role: 'user'), not a system message; the system message is set by the developer before any user interaction to define the assistant's behavior.

Why this answer

How to eliminate wrong answers

Option A is wrong because clustering is an unsupervised learning technique that groups customers into segments based on similarity without a target label, whereas this problem requires a supervised prediction of a specific binary outcome. Option C is wrong because regression predicts a continuous numeric value (e.g., customer lifetime value in dollars), not a discrete binary category like cancel/stay. Option D is wrong because generative AI is used to create new content (e.g., personalized emails), not to perform predictive classification of customer behavior.

Why this answer

How to eliminate wrong answers

Option A is wrong because a model refusing to answer a question is typically a safety or alignment feature (e.g., content filtering or refusal to comply with harmful prompts), not a hallucination. Option C is wrong because processing images instead of text describes a multimodal capability, not a hallucination; hallucinations can occur in text-only models. Option D is wrong because running out of context window space causes truncation or loss of earlier context, leading to incoherence or forgetting, but not the generation of plausible-sounding falsehoods that characterize a hallucination.

Why this answer

How to eliminate wrong answers

Option A is wrong because Optical Character Recognition (OCR) extracts text from images, not background removal. Option C is wrong because Image Captioning generates a human-readable description of the image content, not background manipulation. Option D is wrong because Object Detection identifies and locates objects within an image using bounding boxes, but does not remove or alter the background.

Why this answer

How to eliminate wrong answers

Option A is wrong because temperature controls the randomness or creativity of the output, not the count of completions; it influences token probability distributions but does not determine how many responses are returned. Option B is wrong because max_tokens limits the total number of tokens in a single completion, not the number of completions; it caps response length but does not affect multiplicity. Option D is wrong because stop defines a sequence that halts generation for each completion, used to control output structure, not to specify how many completions are produced.

Why this answer

How to eliminate wrong answers

Option A is wrong because Transparency is about making AI systems understandable and providing clear documentation on how decisions are made, not about providing mechanisms for redress. Option C is wrong because Reliability focuses on the system's ability to perform consistently and correctly under expected conditions, not on user feedback or complaint channels. Option D is wrong because Fairness addresses bias mitigation and equitable treatment across demographic groups, not the process for raising concerns or seeking remedies.

Why this answer

How to eliminate wrong answers

Option B is wrong because unsupervised learning does not use labeled data; it finds hidden patterns or clusters in unlabeled data, which is not the case here as the dataset includes the target outcome. Option C is wrong because reinforcement learning involves an agent learning through trial-and-error interactions with an environment to maximize a reward signal, not from a static labeled dataset. Option D is wrong because semi-supervised learning uses a mix of labeled and unlabeled data, but the question explicitly states the dataset includes labeled outcomes for previous customers, implying all data is labeled.

Why this answer

How to eliminate wrong answers

Option A is wrong because language detection identifies natural human languages (e.g., English, Spanish), not programming languages; Azure AI Language does not include code snippet analysis. Option C is wrong because checking for content that violates community guidelines is a content moderation task, typically performed by Azure AI Content Safety or the Text Analytics for health service, not language detection. Option D is wrong because translating text from any language into a user's preferred language is the function of Azure AI Translator, not language detection, which only identifies the language without performing translation.

Why this answer

How to eliminate wrong answers

Option A is wrong because key phrase extraction identifies important words or phrases in text but does not assign documents to predefined categories. Option C is wrong because sentiment analysis determines the emotional tone (positive, negative, neutral) of text, not topic-based categorization. Option D is wrong because language detection identifies the language of the text (e.g., English, Spanish), which is irrelevant to categorizing support tickets by issue type.

Why this answer

How to eliminate wrong answers

Option B is wrong because increasing the frequency penalty reduces the likelihood of repeating common phrases but does not prevent the model from inventing new, plausible-sounding details; it addresses repetition, not grounding. Option C is wrong because decreasing the temperature to 0 makes the model more deterministic and predictable, but it does not restrict the model to using only provided data; the model can still hallucinate based on its training data. Option D is wrong because enabling content filtering blocks harmful or biased language but does not ensure the model's responses are strictly based on the provided product data; it addresses safety, not factual grounding.

Why this answer

How to eliminate wrong answers

Option A (Fairness) is wrong because the issue is not about biased outcomes or discrimination in recommendations, but about unauthorized data collection and lack of consent. Option B (Reliability and safety) is wrong because the system is not failing or causing physical harm; the concern is about data privacy, not system robustness or safety failures. Option D (Transparency) is wrong because while the lack of informing customers touches on transparency, the primary violation is the unauthorized use of PII, which falls under Privacy and security; transparency is a supporting principle but not the most directly violated one here.

Why this answer

How to eliminate wrong answers

Option A is wrong because generalization is about performance on new data, not about compatibility with programming languages or platforms; Azure ML models can be deployed to various runtimes (e.g., Python, C#, ONNX) but that is a separate deployment concern. Option C is wrong because generalization does not involve generating plain-language descriptions; that describes model interpretability or explainability tools (e.g., Azure ML's model explanations), not the core concept of generalization. Option D is wrong because generalization does not mean a single model works for all tasks; it means the model performs well on unseen data for its specific task, and a model trained for one domain (e.g., image classification) cannot generalize to unrelated tasks (e.g., sentiment analysis).

Why this answer

How to eliminate wrong answers

Option B (R-squared) is wrong because it measures the proportion of variance explained by the model, not the magnitude or penalty of individual errors, so it does not specifically penalize large errors. Option C (Mean Absolute Percentage Error, MAPE) is wrong because it expresses error as a percentage, which can be misleading when actual values are small or zero, and it still averages errors without squaring, so it does not disproportionately penalize large absolute errors. Option D (F1 score) is wrong because it is a classification metric that combines precision and recall, and it is not applicable to regression tasks like house price prediction.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Content Safety is not a network firewall; it analyzes content for harmful material, not network traffic, and does not block malicious packets or use firewall rules. Option C is wrong because it describes a data privacy tool (like Azure Purview or DICOM de-identification), not a content safety service; Content Safety does not detect or redact personal data from training datasets. Option D is wrong because it refers to copyright detection or plagiarism checking, which is not a capability of Azure AI Content Safety; the service focuses on harmful content categories, not intellectual property infringement.

Why this answer

How to eliminate wrong answers

Option A is wrong because Optical Character Recognition (OCR) extracts text from images, not physical objects, so it cannot detect or count shelf items. Option B is wrong because image classification assigns a single label to an entire image (e.g., 'shelf with items'), but it does not provide bounding boxes or a count of individual items. Option D is wrong because semantic segmentation classifies every pixel in an image into a category (e.g., 'shelf', 'item'), but it does not distinguish between individual object instances, making it unsuitable for counting discrete items.

Why this answer

How to eliminate wrong answers

Option A is wrong because reinforcement learning involves an agent learning to make decisions by interacting with an environment to maximize cumulative reward, which is not applicable to grouping static historical purchase data. Option B is wrong because supervised classification requires labeled training data with predefined categories, but the problem explicitly states there are no predefined categories. Option D is wrong because supervised regression predicts a continuous numeric value (e.g., future spending amount) rather than discovering natural groups in data.

Why this answer

How to eliminate wrong answers

Option A is wrong because generating fake personal information is not a purpose of PII detection; that would be data synthesis or anonymization, which is a separate capability. Option C is wrong because verifying user identity is a function of authentication services like Azure Active Directory, not a text analysis feature. Option D is wrong because encrypting stored data is a data protection mechanism handled by services like Azure Key Vault or SQL TDE, not by a natural language processing API.

Why this answer

How to eliminate wrong answers

Option B (Inclusiveness) is wrong because Inclusiveness focuses on designing AI systems that empower and engage everyone, including people with disabilities, but does not specifically mandate the investigation of statistical bias after attribute removal. Option C (Reliability and safety) is wrong because it concerns ensuring the system operates consistently and safely under expected conditions, not addressing fairness or bias in outcomes. Option D (Transparency) is wrong because Transparency is about making AI systems understandable and communicating limitations, not directly requiring the detection and correction of hidden bias.

Why this answer

How to eliminate wrong answers

Option A is wrong because fine-tuning updates the model's weights using the knowledge base, which can cause catastrophic forgetting of other capabilities and does not guarantee that the model will use only the most current information from the knowledge base at inference time. Option B is wrong because zero-shot learning relies entirely on the model's pre-existing internal knowledge without any external retrieval, so the chatbot would not be constrained to the company's specific knowledge base. Option D is wrong because prompt engineering with few-shot examples provides in-context examples but does not dynamically retrieve and inject relevant, up-to-date content from the knowledge base, leaving the model free to generate answers from its internal training data.

Why this answer

How to eliminate wrong answers

Option A (Sentiment Analysis) is wrong because it detects positive, negative, or neutral sentiment in text, not for answering questions from documents. Option B (Key Phrase Extraction) is wrong because it extracts key terms and phrases from text, but does not support a Q&A knowledge base or answer generation. Option D (Language Detection) is wrong because it identifies the language of text, not for building a document-based Q&A system.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Speech focuses on speech-to-text, text-to-speech, and speech translation, not on extracting text from images. Option C is wrong because Azure AI Language provides natural language processing (e.g., sentiment analysis, key phrase extraction) but does not perform OCR or image-based text extraction. Option D is wrong because Azure AI Translator translates text between languages but cannot read or extract text from images.

Why this answer

How to eliminate wrong answers

Option A is wrong because Named Entity Recognition (NER) identifies and categorizes specific entities like people, organizations, locations, and dates, not abstract recurring concepts or themes like 'product quality' or 'shipping delay'. Option C is wrong because Language Detection only identifies the language of the text (e.g., English, Spanish) and provides no insight into the topics or concepts discussed. Option D is wrong because Sentiment Analysis evaluates the emotional tone (positive, negative, neutral) of the text, not the specific topics or recurring concepts mentioned.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure Face API is designed for facial detection, recognition, and analysis, not for extracting text from documents. Option C is wrong because Azure Custom Vision is used for image classification and object detection, not for text extraction from scanned documents. Option D is wrong because Azure Video Indexer is used to extract insights from video content, such as speech and faces, not from static scanned invoices.

Why this answer

How to eliminate wrong answers

Option A is wrong because model lineage is not about the conceptual family tree of architectures or design inspiration; it is a concrete, automated record of the specific resources used in a training run. Option C is wrong because model lineage does not track geographic origins of training data; while data provenance may include location, lineage focuses on the exact datasets, code, and parameters used, not regional collection details. Option D is wrong because model lineage is broader than just deployment version sequences; it encompasses the entire training lifecycle, including the data, code, and compute, not merely the order of production deployments.

Why this answer

How to eliminate wrong answers

Option A is wrong because inclusiveness is about the AI system's impact on users, not about including all team members in development; team composition is a project management concern, not a Responsible AI principle. Option C is wrong because making AI models available regardless of budget relates to affordability or democratization, not inclusiveness; the principle focuses on equitable outcomes for diverse user groups, not organizational access. Option D is wrong because diverse training data is a technique to improve model accuracy and reduce bias, but inclusiveness as a principle is broader, addressing the system's ability to serve all people effectively, not just data diversity.

Why this answer

How to eliminate wrong answers

Option A is wrong because the difference is not about image size; both services can handle images of varying sizes, and Azure AI Vision has specific size limits (e.g., 4 MB for analysis) while Custom Vision also has its own constraints. Option C is wrong because both services can be used in production and development; Custom Vision is often used in development to train a model, then deployed to production, and Azure AI Vision is used in both stages for general analysis. Option D is wrong because the distinction is not about the source of images (camera vs. documents); Azure AI Vision can analyze images from cameras or documents (e.g., OCR on scanned documents), and Custom Vision can be trained on any image type, including document images for custom classification.

Why this answer

How to eliminate wrong answers

Option A is wrong because storing images in Azure Blob Storage with automatic tagging is a separate capability (e.g., using Azure Computer Vision's image tagging or custom vision), not the primary use case of multimodal embeddings for retrieval. Option C is wrong because automatically resizing images for different screen sizes is a media processing task, often handled by Azure Media Services or Content Delivery Network, not by AI Vision's image retrieval. Option D is wrong because detecting copyrighted images in user-uploaded content is typically done with content moderation or fingerprinting services (e.g., Azure Content Moderator or custom hash-based systems), not by multimodal embeddings which focus on semantic similarity search.

Why this answer

How to eliminate wrong answers

Option A is wrong because Prebuilt Text Analytics provides pre-trained models for common tasks like sentiment analysis, key phrase extraction, and language detection, but it does not support custom classification into user-defined categories like 'Billing' or 'Technical'. Option C is wrong because Language Understanding (LUIS) is designed for intent recognition and entity extraction in conversational contexts (e.g., chatbots), not for classifying long-form text documents like support tickets into fixed categories. Option D is wrong because Translator is a machine translation service that converts text between languages and has no capability for text classification or category assignment.

Why this answer

How to eliminate wrong answers

Option A is wrong because adjusting training data labels (e.g., relabeling or correcting mislabeled data) is a data quality or data preprocessing step, not hyperparameter tuning; hyperparameters are settings that govern the training algorithm, not the data itself. Option C is wrong because reducing the number of features (dimensionality reduction or feature selection) is a data preprocessing technique to simplify the model or avoid overfitting, not a search over training configuration settings. Option D is wrong because updating model weights based on new production data describes online learning or model retraining (often via continuous integration/continuous deployment pipelines), not the pre-training search for optimal hyperparameters.

Why this answer

How to eliminate wrong answers

Option A is wrong because fairness is about ensuring the system does not discriminate against groups, but the question focuses on the lack of explanation, not on biased outcomes. Option C is wrong because reliability and safety concern the system's ability to perform consistently and without harm, whereas the issue here is the absence of decision rationale, not system failures or unsafe behavior. Option D is wrong because privacy and security involve protecting personal data and preventing unauthorized access, not providing explanations for decisions.

Why this answer

How to eliminate wrong answers

Option A is wrong because AI is not a type of computer hardware; it is a software discipline that can run on various hardware, including CPUs, GPUs, and TPUs, but the hardware itself is not AI. Option C is wrong because AI is not a programming language; languages like Python, R, or C++ are used to implement AI algorithms, but the concept of AI is independent of any specific language. Option D is wrong because AI is not a database; while AI systems often use databases to store training data or results, the core of AI is the algorithms and models that process and learn from that data, not the storage mechanism.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes bibliographic citation validation, which is a document formatting task unrelated to NLP entity resolution. Option C is wrong because it describes language identification or code-switching handling, not the resolution of referring expressions within a single language. Option D is wrong because it describes a database referential integrity concept (foreign key to primary key matching), which is a data management operation, not a natural language processing task.

Why this answer

How to eliminate wrong answers

Why this answer

How to eliminate wrong answers

Option A is wrong because Image Classification assigns a single label to the entire image (e.g., 'defective' or 'non-defective'), but it cannot identify the location or shape of individual defects. Option B is wrong because Object Detection draws bounding boxes around objects, which is too coarse for irregularly shaped defects like scratches or cracks that require pixel-accurate boundaries. Option D is wrong because Dense Captioning generates descriptive captions for image regions, but it does not produce a pixel-level segmentation map needed to precisely outline defect shapes.

Why this answer

How to eliminate wrong answers

Option A is wrong because grammatically correcting and editing text for quality is a feature of language models like GPT or grammar checkers, not Azure AI Content Safety's text moderation, which focuses on harmful content detection. Option C is wrong because moderating the volume of text to prevent spam is typically handled by rate limiting or spam filters, not by Azure AI Content Safety's NLP-based content analysis. Option D is wrong because identifying and removing personally identifiable information (PII) is a task for Azure AI Language's PII detection or Azure Purview, not for text moderation in Content Safety, which targets offensive or harmful content.

Why this answer

How to eliminate wrong answers

Option A is wrong because binary classification predicts one of two discrete categories (e.g., yes/no), not a continuous number like units sold. Option B is wrong because multiclass classification predicts one of three or more discrete classes (e.g., product categories), not a continuous value. Option D is wrong because clustering groups unlabeled data into clusters based on similarity, without predicting a specific numeric target; it is an unsupervised learning task, not a supervised prediction of a continuous variable.

Why this answer

How to eliminate wrong answers

Option B (Fairness) is wrong because it addresses bias and equitable treatment across demographic groups, not the physical stability of HVAC output. Option C (Transparency) is wrong because it concerns explainability and user understanding of AI decisions, not the system's operational safety or reliability. Option D (Privacy and security) is wrong because it focuses on data protection and unauthorized access, not the prevention of temperature swings or physical discomfort.

Why this answer

How to eliminate wrong answers

Option A is wrong because 'Speed, Accuracy, Cost, Scalability, Security, Compliance' are performance and operational metrics for IT systems, not the ethical pillars of Microsoft's Responsible AI framework. Option C is wrong because 'Innovation, Efficiency, Quality, Agility, Trust, Sustainability' are generic business or agile development principles, not the specific six pillars defined by Microsoft for responsible AI. Option D is wrong because 'Openness, Collaboration, Transparency, Community, Excellence, Impact' are values common in open-source or community-driven projects, but they do not match Microsoft's official Responsible AI pillars, which include Fairness, Reliability & Safety, Privacy & Security, Inclusiveness, Transparency, and Accountability.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes authentication (verifying credentials), which is a security function handled by Azure Active Directory or identity platforms, not by the Azure AI Language service. Option C is wrong because it describes auditing access to Azure resources (e.g., via Azure Monitor or Activity Logs), not detecting PII in text content. Option D is wrong because it implies contextual judgment about inappropriate sharing, which is a subjective policy decision; Azure AI Language's PII detection identifies entities based on patterns and models, not on the appropriateness of the context.

Why this answer

How to eliminate wrong answers

Option A is wrong because underfitting would show poor performance on both training and test sets, not a large gap between them. Option C is wrong because data leakage would typically cause artificially high performance on both sets, not a significant drop on the test set. Option D is wrong because multicollinearity affects coefficient stability and interpretability but does not inherently cause a large training-test performance gap; it can exist even in a well-generalized model.