Microsoft Azure AI Fundamentals AI-900 (AI-900) — Questions 376–450

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes time-based or calendar-based maintenance, which is a traditional, non-AI approach that does not use any predictive models or data-driven insights. Option C is wrong because it refers to model maintenance (retraining to preserve accuracy), which is an MLOps activity, not a workload that predicts equipment failures. Option D is wrong because it describes automated software bug fixing, which is a different AI domain (e.g., program repair or self-healing systems) and has nothing to do with predicting physical equipment failures.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Language Studio is specifically designed for NLP workloads, not for training custom computer vision models; that purpose is served by Azure AI Custom Vision or Azure AI Vision Studio. Option C is wrong because billing management for Azure AI services is handled through the Azure Portal's Cost Management + Billing blade, not through Language Studio. Option D is wrong because Language Studio is a no-code UI, not a code editor; writing Python ML scripts is done in tools like Visual Studio Code, Jupyter Notebooks, or Azure Machine Learning studio.

Why this answer

How to eliminate wrong answers

Option A is wrong because pre-built entity recognition only recognizes common, generic entity types (e.g., person, location, organization) and cannot be customized to extract domain-specific categories like symptoms or medication names. Option B is wrong because key phrase extraction returns a list of key talking points or important phrases from text, but it does not classify or extract specific entities into pre-defined categories. Option C is wrong because custom text classification assigns a category or label to an entire document or sentence, not to specific text spans within the text.

Why this answer

How to eliminate wrong answers

Option A is wrong because manually customising a visual theme or layout is a static user preference setting, not an AI-driven workload that adapts in real time based on behaviour. Option C is wrong because recommending specific items based on purchase history is a classic recommendation system, which is a subset of personalisation but does not encompass the full dynamic adaptation of the user experience. Option D is wrong because creating personalised data privacy policies based on location is a compliance or policy automation task, not an AI workload focused on adapting the user experience.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure AI Studio is not a video editing tool; it is a platform for developing AI applications, not for media editing. Option C is wrong because Azure AI Studio is not limited to Python machine learning code; it supports multiple languages and includes visual tools for building AI workflows, not just an IDE. Option D is wrong because Azure AI Studio is not a database service; it can integrate with databases like Azure Cosmos DB for storing conversation history, but it is not a database service itself.

Why this answer

How to eliminate wrong answers

Option A is wrong because prompt engineering only involves crafting the input prompt to guide the model's behavior, but it does not inject external, up-to-date documents as context; it relies solely on the model's pre-existing knowledge. Option B is wrong because fine-tuning would retrain the model on policy documents, which is a more resource-intensive process that updates the model's parameters, whereas the scenario describes providing extra context at inference time without altering the model. Option D is wrong because a content filter is a post-processing safety mechanism that blocks or flags harmful or inappropriate outputs, not a technique to supply factual context for accuracy.

Why this answer

Why this answer

How to eliminate wrong answers

Option B (unsupervised learning) is wrong because it is used for finding hidden patterns or groupings in unlabeled data, such as clustering, and does not use labeled examples to predict a specific category like 'cat' or 'dog'. Option C (reinforcement learning) is wrong because it involves an agent learning to make sequential decisions by interacting with an environment and receiving rewards or penalties, which is not applicable to static image classification tasks. Option D (semi-supervised learning) is wrong because it combines a small amount of labeled data with a large amount of unlabeled data; here the dataset already contains thousands of labeled images, so there is no need to leverage unlabeled data, and the problem is fully supervised.

Why this answer

How to eliminate wrong answers

Option A is wrong because Sentiment Analysis determines the emotional tone (positive, negative, neutral) of text, not the extraction of specific issue-related phrases. Option C is wrong because Named Entity Recognition (NER) identifies and categorizes named entities like people, organizations, locations, and dates, not general descriptive phrases like 'slow internet connection'. Option D is wrong because Language Detection identifies the language of the text (e.g., English, Spanish), which is irrelevant to extracting customer-reported issues.

Why this answer

How to eliminate wrong answers

Option A is wrong because the Phi models are not from OpenAI; they are Microsoft's own small language models, and they are not designed for the highest capability complex tasks—that role belongs to large models like GPT-4. Option C is wrong because the Phi models are general-purpose language models, not specialized for structured financial data; they handle natural language across domains. Option D is wrong because the Phi models are text-based language models, not image generation models; they do not compete with DALL-E for artistic content creation.

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') and cannot locate multiple defects or their positions. Option C is wrong because Optical Character Recognition (OCR) extracts text from images, not visual defects like scratches or dents. Option D is wrong because Face Detection identifies human faces, not product defects.

Why this answer

How to eliminate wrong answers

Option A is wrong because fairness addresses bias and equitable treatment across groups (e.g., ensuring the model does not discriminate by race or gender), not the prevention of individual identification. Option B is wrong because reliability and safety focus on the system performing consistently and without causing harm (e.g., accuracy under varied conditions), not on data privacy. Option D is wrong because inclusiveness ensures the system serves diverse user groups (e.g., supporting multiple languages or accessibility), which is unrelated to anonymizing training data.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes performance tests for the web interface loading speed, which is unrelated to model benchmarks; model benchmarks evaluate AI task performance, not UI responsiveness. Option C is wrong because it refers to Azure's SLA guarantees for model availability and API response time, which are operational metrics, not performance benchmarks for model selection. Option D is wrong because it describes pricing comparisons against competitor services, which is a cost analysis, not a performance benchmark for AI tasks.

Why this answer

How to eliminate wrong answers

Option A is wrong because it incorrectly reduces responsible AI to a narrow legal compliance tactic of using only 'Azure-approved models,' whereas the actual practice involves a broad set of principles and tools applied across the entire lifecycle, not just model selection. Option C is wrong because it misrepresents responsible AI as a restrictive design that limits responses to pre-approved questions, which contradicts the goal of building flexible, transparent, and safe generative AI systems that can handle diverse inputs while being monitored for harmful outputs. Option D is wrong because it overemphasizes a single bureaucratic step (legal review before every deployment) rather than the continuous, integrated process of embedding ethical checks and safety tools throughout design, development, and operations.

Why this answer

How to eliminate wrong answers

Option A is wrong because object detection and image classification do not produce the same output; classification labels the entire image, while detection outputs bounding boxes and labels for each object. Option C is wrong because while image classification is generally simpler and can be faster, the statement is not a defining difference—object detection is not inherently slower in all implementations, and the question asks for the functional difference, not performance. Option D is wrong because object detection is specifically designed to handle multiple objects in a single image, not just a single object; classification can also handle multiple objects but only produces one label for the whole scene.

Why this answer

How to eliminate wrong answers

Option A is wrong because frequency_penalty reduces repetition by penalizing tokens that have already appeared in the text, which decreases diversity rather than increasing it. Option B is wrong because presence_penalty encourages the model to talk about new topics by penalizing tokens that have appeared at all, but it does not directly control the randomness or creativity of word choices. Option D is wrong because top_p (nucleus sampling) limits token selection to a cumulative probability mass (e.g., 0.9), which can reduce diversity by cutting off the long tail of low-probability tokens, whereas the team wants more unexpected choices.

Why this answer

How to eliminate wrong answers

Option B is wrong because Custom Text Classification requires you to provide labeled training data and train a custom model, which is unnecessary overhead when prebuilt Text Analytics already handles key phrase extraction and sentiment analysis out-of-the-box. Option C is wrong because Conversational Language Understanding (CLU) is designed for intent recognition and entity extraction in conversational flows (e.g., for bots), not for extracting key phrases or performing sentiment analysis on chat transcripts. Option D is wrong because Question Answering is used to provide answers from a knowledge base or FAQ, not for extracting product issues or analyzing sentiment.

Why this answer

How to eliminate wrong answers

Option A is wrong because model export does not involve exporting training logs or metrics to Excel; those are accessed via training APIs or the Azure portal for analysis, not as an export feature. Option C is wrong because exporting the training data to another Azure service is not a built-in Custom Vision feature; data can be exported manually, but the 'model export' feature specifically exports the trained model artifact, not the dataset. Option D is wrong because Custom Vision does not export projects as YAML configuration files; project configuration is managed through the portal or SDK, and YAML exports are not a supported format for model deployment.

Why this answer

How to eliminate wrong answers

Option A is wrong because healthcare NLP does not perform translation between languages; translation is handled by the Azure Translator service, not by the healthcare NLP models. Option C is wrong because healthcare NLP extracts information from unstructured text, not from structured EHR data, and it does not generate reports—report generation would require additional services like Azure Logic Apps or Power BI. Option D is wrong because healthcare NLP does not diagnose conditions; it only extracts and normalizes medical entities from text, leaving diagnosis to clinical judgment or separate AI models.

Why this answer

How to eliminate wrong answers

Option A is wrong because generating visual layouts for new document templates is not a capability of the layout model; it is an extraction tool, not a design tool. Option C is wrong because converting documents between file formats (e.g., PDF to DOCX) is not a function of the layout model; format conversion is handled by separate document processing libraries or services. Option D is wrong because checking for grammatical and spelling errors falls under natural language processing (NLP) services like Azure AI Language, not the layout model, which focuses on spatial and structural extraction.

Why this answer

How to eliminate wrong answers

Option B (OCR) is wrong because OCR extracts text from images (e.g., product labels or barcodes) but does not detect or count physical objects like cereal boxes; it cannot learn to recognize 'Brand A cereal' as a visual object without text. Option C (Prebuilt image analysis - Describe Image) is wrong because it generates a human-readable caption of the scene (e.g., 'a shelf with boxes') and provides generic tags, but it cannot be trained on custom product categories or output per-product counts with bounding boxes. Option D (Facial recognition) is wrong because it is designed to detect, analyze, and verify human faces, not inanimate objects like cereal boxes on a shelf.

Why this answer

How to eliminate wrong answers

Option A is wrong because MAE does not measure confidence percentages; classification confidence is typically evaluated using metrics like log loss or calibration curves. Option C is wrong because MAE averages all absolute errors without applying a threshold; metrics like 'accuracy within a tolerance' or 'pinball loss' handle threshold-based deviations. Option D is wrong because MAE compares predictions to actual values, not to random chance; comparing to random chance is done with metrics like R-squared or relative absolute error.

Why this answer

How to eliminate wrong answers

Option B is wrong because precision for the default class would be undefined (0/0) or 0% if the model never predicts default, but it does not directly show that the model misses all actual defaults—it focuses on the accuracy of positive predictions. Option C is wrong because the F1 score is the harmonic mean of precision and recall; if recall is 0%, the F1 score is also 0%, but it is a composite metric that does not isolate the failure to identify defaults as clearly as recall. Option D is wrong because accuracy is 98% due to the class imbalance, masking the model's complete failure on the minority class; accuracy alone cannot reveal the lack of true positive predictions.

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 multiple defect types or their locations within the same image. Option C is wrong because semantic segmentation assigns a class label to every pixel, which is overkill for defect type identification and does not inherently separate individual defect instances or provide bounding boxes. Option D is wrong because optical character recognition (OCR) extracts text from images, which is irrelevant to detecting physical defects like scratches, cracks, or dents.

Why these pairings

Why this answer

How to eliminate wrong answers

Option A is wrong because sentiment analysis determines the emotional tone (positive, negative, neutral) of text, not the main topics or themes discussed. Option C is wrong because named entity recognition identifies specific entities like people, organizations, and locations, but does not extract general topic labels or themes. Option D is wrong because language detection identifies the language of the text (e.g., English, Spanish), not the topics or subject matter within the document.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes camera control software (e.g., drivers or APIs for adjusting exposure, focus, or capture), which is a hardware management task, not an AI workload. Option C is wrong because it describes generative AI (e.g., text-to-image models like DALL-E), which creates new images rather than recognizing content in existing ones. Option D is wrong because it describes a database retrieval operation (e.g., using a key-value store or blob storage with unique IDs), which is a data management function, not an AI capability.

Why this answer

How to eliminate wrong answers

Option A is wrong because AI systems do not automatically fix their own errors; accountability requires human oversight to identify and correct errors, not autonomous self-healing. Option C is wrong because while logging user interactions can support auditing, it is a specific technical practice, not the core definition of accountability, which focuses on human responsibility and governance. Option D is wrong because accountability does not mandate open-source code; it requires transparency and explainability, but code can remain proprietary as long as humans are accountable for the system's behavior.

Why this answer

How to eliminate wrong answers

Option A is wrong because Language Detection identifies the language of the text (e.g., English, Spanish), not the topics or key phrases within the message. Option C is wrong because Sentiment Analysis determines the emotional tone (positive, negative, neutral) of the text, not the specific topics mentioned. Option D is wrong because Entity Recognition extracts named entities like people, places, and organizations, but does not extract multi-word phrases or abstract concepts like 'shipping delay' or 'refund policy'.

Why this answer

How to eliminate wrong answers

Option A is wrong because a spelling checker is a classic example of Natural Language Processing (NLP) and pattern matching, not a recommendation system; it corrects text based on dictionary rules or language models, not by predicting user preferences. Option C is wrong because a database storing customer purchase history is simply a data storage and retrieval system; it does not involve any AI model or algorithm to generate recommendations—it is a prerequisite for a recommendation system but not the AI workload itself. Option D is wrong because a barcode scanner is a pure input/output hardware device that reads optical codes; it involves no AI inference or learning, and is typically associated with Computer Vision only if it uses image recognition, but here it is a simple scanner.

Why this answer

How to eliminate wrong answers

Option A is wrong because regression is a supervised learning technique used for predicting continuous numerical values (e.g., sales amount), not for discovering natural groupings. Option B is wrong because classification is a supervised learning method that requires labeled training data to assign predefined categories, whereas the dataset has no predefined categories. Option D is wrong because reinforcement learning involves an agent learning optimal actions through trial-and-error interactions with an environment to maximize cumulative reward, which is unrelated to grouping customers based on historical data.

Why this answer

How to eliminate wrong answers

Option A is wrong because accuracy measures overall correct predictions (true positives + true negatives divided by total predictions) and can be misleading in imbalanced datasets like fraud detection, where legitimate transactions vastly outnumber fraudulent ones; a model that always predicts 'not fraud' could achieve high accuracy but fail to catch any fraud. Option B is wrong because precision focuses solely on the proportion of flagged transactions that are actually fraudulent (true positives / (true positives + false positives)), which minimizes false positives but does not account for missed fraudulent transactions (false negatives), potentially allowing many frauds to go undetected. Option C is wrong because recall (sensitivity) measures the proportion of actual fraudulent transactions correctly identified (true positives / (true positives + false negatives)), which prioritizes catching fraud but can lead to a high number of false positives, directly conflicting with the business goal of minimizing customer dissatisfaction.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure Computer Vision Image Analysis provides pre-built models for general image tagging, OCR, and object detection, but it cannot be trained on custom datasets to recognize specific bird species. Option C is wrong because Azure Face API is designed specifically for detecting and recognizing human faces, not animals or birds. Option D is wrong because Azure Form Recognizer is optimized for extracting text and structured data from documents like invoices and forms, not for image classification of natural subjects.

Why this answer

How to eliminate wrong answers

Option A is wrong because the API endpoint URL is simply the address used to send requests to the Azure OpenAI service, not a feature for calling external functions. Option C is wrong because calling Azure support is a customer service action, not a technical capability of the AI model. Option D is wrong because tool calling is not a billing mechanism; billing is based on token usage and API calls, but the feature itself is about enabling external function invocation, not counting calls.

Why this answer

How to eliminate wrong answers

Option A (Object detection) is wrong because it returns bounding boxes around objects, not pixel-level outlines, so it cannot draw precise outlines around each equipment type. Option C (Image classification) is wrong because it assigns a single label to the entire image, failing to identify multiple equipment types or their locations. Option D (Optical Character Recognition) is wrong because it extracts text from images, not heavy equipment, and has no capability for object localization or segmentation.

Why this answer

How to eliminate wrong answers

Option A is wrong because GPT-4o is not text-only; it is multimodal, and while it may offer speed improvements, its defining feature is native multimodal processing, not just faster text responses. Option C is wrong because GPT-4o is not specialized for code generation; it is a general-purpose multimodal model, though it can generate code as part of its capabilities. Option D is wrong because GPT-4o is not an older or less capable version; it is a newer, more advanced model that extends GPT-4 with native multimodal support.

Why this answer

How to eliminate wrong answers

Option A is wrong because Privacy concerns the secure handling of personal data (e.g., GDPR compliance, data encryption), not the equitable treatment of users across different backgrounds. Option B is wrong because Inclusiveness focuses on designing AI to empower and engage people of all abilities (e.g., accessibility features like speech-to-text for hearing impairments), not on preventing unfair bias based on demographics. Option D is wrong because Transparency involves making AI decisions understandable and auditable (e.g., model interpretability with SHAP values), which supports fairness but does not directly address the requirement of treating all users fairly.

Why this answer

How to eliminate wrong answers

Option A is wrong because it refers to a generic or non-existent feature; Azure AI Language does not have a feature labeled 'A' for custom classification. Option B is wrong because it likely represents a pre-built model like Sentiment Analysis or Key Phrase Extraction, which cannot be trained on custom categories and would not classify tickets into predefined topics like 'billing' or 'technical issue'. Option D is wrong because it probably represents a feature like Language Detection or Entity Recognition, which are not designed for custom multi-class classification tasks and require no labeled training data.

Why these pairings

Why this answer

How to eliminate wrong answers

Option A is wrong because object detection identifies and locates objects (e.g., packages, boxes) within an image by drawing bounding boxes, but it does not extract text or read serial numbers. Option C is wrong because image classification assigns a single label or category to an entire image (e.g., 'package' or 'label'), but it cannot read or extract specific alphanumeric strings like serial numbers. Option D is wrong because facial recognition is designed to detect, analyze, and identify human faces, not to process text on labels or packages.

Why this answer

How to eliminate wrong answers

Option A is wrong because Azure OpenAI Service is not for building traditional rule-based chatbots; it uses large language models for generative AI, not predefined rules. Option C is wrong because it is not a machine learning training platform for custom models; it provides pre-trained OpenAI models via API, not a service to train your own models from scratch. Option D is wrong because it is not a database service for storing AI training data; it is an AI inference service, and Azure offers separate services like Azure Cosmos DB or Azure Blob Storage for data storage.

Why this answer

How to eliminate wrong answers

Option B (Object detection) is wrong because it identifies and locates objects (e.g., cars, animals) within an image, not text characters or words. Option C (Image classification) is wrong because it assigns a single label or category to an entire image (e.g., 'book cover'), rather than extracting specific text content. Option D (Facial detection) is wrong because it detects human faces and their attributes (e.g., age, emotion), which is irrelevant to converting printed text into editable format.

Why this answer

How to eliminate wrong answers

Option A is wrong because fine-tuning does not repair errors in the base training data; it adapts the model to new data, and any errors in the original training data would require retraining from scratch or data correction. Option C is wrong because adjusting the temperature setting is a hyperparameter tuning technique for controlling output randomness, not a training process that modifies the model's weights. Option D is wrong because selecting a pre-trained model from the Azure model catalogue is a model selection step, not a training or adaptation process like fine-tuning.

Why this answer

How to eliminate wrong answers

Option B (Image classification) is wrong because it assigns a single label to the entire image, not identifying or locating individual objects. Option C (Semantic segmentation) is wrong because it classifies every pixel into a category (e.g., 'box' vs 'shelf') but does not separate individual instances of the same class, so it cannot draw distinct bounding boxes around each box. Option D (Optical Character Recognition) is wrong because it extracts text from images, not relevant to locating physical boxes.

Why this answer

How to eliminate wrong answers

Option A is wrong because Fairness focuses on ensuring AI does not discriminate against groups (e.g., based on gender or race), not on explaining individual decisions. Option B is wrong because Reliability and safety concerns the system's consistent performance and error handling, not the interpretability of its outputs. Option D is wrong because Accountability deals with assigning responsibility for the system's behavior and outcomes, not with providing explanations for specific recommendations.

Why this answer

How to eliminate wrong answers

Option A (temperature) is wrong because it controls the randomness of the output, not its length; a lower temperature makes the model more deterministic, but does not cap token count. Option B (top_p) is wrong because it implements nucleus sampling, limiting the cumulative probability of token choices to influence diversity, not the total number of tokens generated. Option C (frequency_penalty) is wrong because it reduces repetition by penalizing tokens that have already appeared, which can affect content but does not enforce a strict length limit.

Why this answer

How to eliminate wrong answers

Option A is wrong because the confidence score is not the percentage of training images containing that object; that would be a class distribution metric, not a per-detection certainty. Option C is wrong because the confidence score is a per-prediction value, not the overall model accuracy measured on a test dataset; test accuracy is a separate evaluation metric. Option D is wrong because the confidence score is computed by the model algorithmically, not assigned by human reviewers; human review is a separate validation step.

Why this answer

How to eliminate wrong answers

Option A is wrong because feature scaling (e.g., standardization or normalization) adjusts the range of feature values but does not reduce the number of features; it is often a preprocessing step before applying PCA, not a dimensionality reduction technique itself. Option C is wrong because regularization (e.g., L1 or L2) penalizes model coefficients to prevent overfitting but does not reduce the number of features in the dataset; it works during model training, not as a preprocessing step. Option D is wrong because cross-validation is a model evaluation technique used to assess generalization performance by splitting data into training and validation folds; it does not reduce feature count or variance retention.

Why this answer

How to eliminate wrong answers

Option A (Fairness) is wrong because fairness addresses bias and equitable treatment across groups, not the protection of individual identities from exposure. Option B (Reliability and safety) is wrong because reliability and safety focus on system accuracy, robustness, and avoiding harmful failures, not on preventing identity disclosure. Option D (Inclusiveness) is wrong because inclusiveness ensures the system benefits diverse user groups and does not exclude or disadvantage anyone, which is unrelated to safeguarding personal identity in reporting.

Why this answer

How to eliminate wrong answers

Option A is wrong because sentiment analysis determines the overall positive, negative, or neutral emotional tone of text, not the specific topics or aspects mentioned. Option C is wrong because named entity recognition identifies and categorizes named entities like people, organizations, locations, and dates, not general descriptive phrases like 'battery life' or 'price'. Option D is wrong because language detection identifies the language in which the text is written (e.g., English, Spanish), not the topics or aspects discussed within the text.

Why this order

Why this answer

How to eliminate wrong answers

Option A is wrong because the maximum number of API requests per second is governed by a separate rate limit (e.g., tokens per minute or requests per minute), not by the 'max tokens' parameter. Option C is wrong because the input prompt length is controlled by the model's context window (e.g., 4096 tokens for GPT-3.5), not by 'max tokens', which only applies to the generated output. Option D is wrong because concurrent user limits are managed through Azure subscription quotas and throughput settings (e.g., provisioned throughput units), not by the 'max tokens' parameter.

Why this answer

How to eliminate wrong answers

Option A is wrong because GPT-4 is a text-based language model, not an image generation model, and its standard content filtering applies to text outputs, not images. Option C is wrong because GPT-3.5 is also a text-only model and cannot generate images; custom moderation would require additional services and does not provide built-in image safety filtering. Option D is wrong because Codex is a model specialized for code generation, not image generation, and output validation is a generic concept, not a specific feature for blocking inappropriate image content.

Why this answer

How to eliminate wrong answers

Option A is wrong because Named Entity Recognition (NER) identifies and extracts specific entities (e.g., people, organizations, dates) from text, but it does not classify entire documents into predefined categories. Option B is wrong because key phrase extraction returns a list of important phrases from the text, but it does not assign a single category label to the document. Option D is wrong because Conversational Language Understanding (CLU) is designed for intent classification and entity extraction in conversational contexts (e.g., chatbots), not for categorizing static legal documents into predefined single-label categories.

Why this answer

How to eliminate wrong answers

Option A is wrong because ONNX is not a programming language; it is a serialized model format, and Azure-specific ML languages include languages like Python or R, not ONNX. Option C is wrong because ONNX does not store training data; it stores model architecture and weights, while databases like Azure SQL or Cosmos DB are used for data storage. Option D is wrong because ONNX is not a cloud service for distributed training; Azure offers services like Azure Machine Learning for distributed training, while ONNX is purely an interchange format.

Why this answer

How to eliminate wrong answers

Option A is wrong because Custom Question Answering is designed for providing predefined answers to user questions based on a knowledge base, not for extracting intents and entities from utterances. Option C is wrong because Custom Text Classification only assigns predefined labels (categories) to entire documents or sentences, without extracting specific entities like destination or date. Option D is wrong because Named Entity Recognition (NER) is a prebuilt feature that extracts common entities (e.g., person, location) but cannot be trained on custom intents or small labeled datasets for domain-specific scenarios like flight booking.

Why this answer

How to eliminate wrong answers

Option A is wrong because Accountability refers to the responsibility of the organization for the AI system's outcomes, not specifically to preventing discrimination in decision-making. Option B is wrong because Inclusiveness focuses on designing AI systems that empower and engage diverse users, but it does not directly address the algorithmic bias or disparate impact in automated loan approvals. Option D is wrong because Reliability and Safety ensures the system operates correctly and safely under expected conditions, but it does not inherently prevent discriminatory outcomes against specific groups.

Why this answer

How to eliminate wrong answers

Option A is wrong because key phrase extraction identifies the main topics or themes in text (e.g., 'surgery', 'recovery'), not specific PII entities like names or SSNs. Option C is wrong because sentiment analysis determines the emotional tone (positive, negative, neutral) of text, which has no capability to detect or remove personal identifiers. Option D is wrong because language detection identifies the language of the text (e.g., English, Spanish), not any form of personally identifiable information.

Why this answer

How to eliminate wrong answers

Option A is wrong because AI models, especially deep neural networks, often require significant computational resources and can be slower at inference than a simple conventional program performing the same deterministic task; speed is not a defining advantage of AI. Option C is wrong because AI is not limited to images; it works with various data types including text, audio, tabular data, and time-series, while conventional programming also works with all data types. Option D is wrong because conventional programming is not inherently more accurate; AI can achieve higher accuracy on tasks like fraud detection or medical diagnosis by learning complex patterns from data, whereas conventional programs are limited by the precision of manually written rules.

Why this answer

How to eliminate wrong answers

Option A is wrong because the standard Text Analytics API does not have the medical domain knowledge to extract conditions, medications, or dosages; it only handles general entity recognition, sentiment, and key phrases. Option C is wrong because Translator Text is for language translation, not transcription, and Language Understanding (LUIS) is for intent and entity extraction from user utterances in conversational AI, not for medical entity extraction from transcribed speech. Option D is wrong because Speaker Recognition identifies who is speaking, not what is said, and Question Answering provides answers from a knowledge base, neither of which transcribes speech or extracts medical entities.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes a quiz application where users answer questions about images, which is a human-driven activity, not an AI system that itself answers questions about images. Option C is wrong because it describes a human interview process with video recording, which has no relation to AI answering questions about image content. Option D is wrong because it describes image generation from prompts (text-to-image or image-to-image), which is the reverse direction of VQA—VQA takes an image and a question to produce an answer, not generate an image.

Why this answer

How to eliminate wrong answers

Option A is wrong because underfitting would result in high error on both training and test data, not low training error and high test error. Option C is wrong because data leakage involves using information from the test set during training, which would artificially inflate training performance but is not described here; the issue is generalization to a new city, not a data contamination problem. Option D is wrong because while bias-variance tradeoff is a related concept, it does not specifically name the phenomenon; overfitting is the direct explanation for low training error and high test error on new data.

Why this answer

How to eliminate wrong answers

Option A is wrong because Optical Character Recognition (OCR) extracts text from images, not descriptions of visual content, so it cannot describe a photo of a landscape or object. Option C is wrong because Object Detection identifies and locates specific objects within an image but does not generate a coherent textual description of the overall scene. Option D is wrong because Face Detection only identifies human faces and their attributes, ignoring other image content and context needed for alternative text.

Why this answer

How to eliminate wrong answers

Option A is wrong because the receipt model is not limited to digital PDFs with standardized formatting; it can process scanned images, photos, and various receipt layouts, including those with non-standard formatting. Option C is wrong because the receipt model is not specialized for medical receipts or prescription records; those would require a different prebuilt model (e.g., the 'health insurance' or custom model). Option D is wrong because electronic bank transfer receipts are not the target of this model; the receipt model focuses on point-of-sale receipts, not financial transaction records from banking systems.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes a human workflow (multiple labelers), not a machine learning paradigm; semi-supervised learning refers to the data labeling strategy, not the number of human supervisors. Option C is wrong because it describes online learning or reinforcement learning, where the model updates from live user feedback, not the semi-supervised combination of labeled and unlabeled data. Option D is wrong because it describes checkpointing or resumable training, which is a fault-tolerance mechanism, not a learning paradigm.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes a licensing or multi-tenancy scenario, not the ethical risk of technology being repurposed for harm. Option C is wrong because using the same model for training and inference is a cost-saving or architectural choice (e.g., in online learning), not a dual-use concern. Option D is wrong because combining two models (ensemble learning) is a technique to improve accuracy, not a problem related to misuse of AI capabilities.

Why this answer

How to eliminate wrong answers

Option A is wrong because fairness addresses bias and equitable treatment across groups, not system robustness to sensor failures. Option C is wrong because transparency focuses on explainability and interpretability of AI decisions, not on maintaining correct operation under faulty inputs. Option D is wrong because accountability concerns assignment of responsibility and governance, not the technical resilience of the system to noisy data.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes a physical file management error (moving files to the wrong storage container), which is an operational issue unrelated to the statistical concept of data drift in model performance. Option C is wrong because moving data between Azure regions for latency optimization is a data replication or caching strategy, not a change in data distribution that degrades model accuracy. Option D is wrong because intentional modification of training data to improve robustness is a data augmentation technique used during training, not a post-deployment concern about naturally occurring shifts in production data.

Why this answer

How to eliminate wrong answers

Option A is wrong because extracting text from documents and images is the function of Azure AI Vision's OCR (Optical Character Recognition) capability, not spatial analysis. Option C is wrong because spatial analysis operates on video feeds from physical cameras, not satellite imagery, and it does not identify 3D coordinates of objects in such imagery. Option D is wrong because generating 3D models from 2D photographs is not a feature of spatial analysis; that would relate to photogrammetry or 3D reconstruction services, not Azure's spatial analysis.

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 an entire image but does not provide bounding boxes or locations for multiple objects, so it cannot detect pedestrians' positions. Option B is wrong because Semantic Segmentation classifies every pixel into a category (e.g., road, pedestrian) but does not extract text from signs, and OCR alone cannot detect pedestrians. Option D is wrong because Face Detection specifically identifies human faces, not full pedestrian bodies, and cannot detect pedestrians crossing the road or read speed limit text.

Why this answer

How to eliminate wrong answers

Option A is wrong because it describes general text formatting (numbered sections, bullet points) which is not JSON mode; JSON mode enforces a specific data structure, not visual layout. Option C is wrong because it describes automatic database persistence, which is not a feature of Azure OpenAI's API—structured output only ensures the response is valid JSON, not that it is saved anywhere. Option D is wrong because JSON mode does not handle multilingual generation; it only constrains the format of the output to JSON, regardless of language.

Why this answer

How to eliminate wrong answers

Option A is wrong because a batch prediction job that scores new data against a deployed model is an inference or scoring job, not a training job. Option C is wrong because a scheduled report on model performance in production is a monitoring or evaluation task, not a training job. Option D is wrong because a data preprocessing pipeline that cleans raw datasets is a data preparation step, which may precede training but is not itself a training job.

Why this answer

How to eliminate wrong answers

Option B is wrong because while image classification can be computationally simpler, the statement that object detection is 'slower but more accurate' is misleading—accuracy depends on the specific model and use case, not a general trade-off; object detection provides more detailed output (locations), not inherently higher accuracy. Option C is wrong because both image classification and object detection can work on videos (e.g., frame-by-frame analysis) and static images; there is no restriction that classification is for videos and detection only for static images. Option D is wrong because image classification and object detection are fundamentally different tasks—classification labels the whole image, while detection identifies and localizes multiple objects, so they are not the same task with different names.

Why this answer

How to eliminate wrong answers

Option A is wrong because 'sentiment analysis at scale' does not mean restricting analysis to only the largest datasets; it means handling any volume of text efficiently, and accuracy is not inherently maximized by dataset size alone. Option C is wrong because scaling the sentiment score range (e.g., from 0-1 to 1-5) is not a feature of Azure AI Language's sentiment analysis; the service returns scores between 0 and 1 for positive, neutral, and negative sentiments, and does not adjust ranges to match external rating systems. Option D is wrong because while Azure AI Language uses sophisticated models, 'sentiment analysis at scale' is about throughput and resource management (e.g., batch processing, rate limiting), not simply using larger models; the service already uses optimized models, and scaling is achieved via API design, not model size.