Machine Learning Fundamentals

Question 1

What is Artificial Intelligence (AI)?

Answer 1

AI refers to when machines, such as computers, can perform tasks that typically require human intelligence, including making predictions, recognizing patterns, understanding and generating human language, and interpreting visual information.

Question 2

What is Machine Learning (ML)?

Answer 2

ML is the foundation of an AI system that can learn and predict. ML models get better at a task without explicit programming, training themselves based on past data to make future predictions.

Question 3

What is Computer Vision (CV)?

Answer 3

CV enables machines to ‘see’ and understand digital images and videos like a human, recognizing patterns, objects, faces, and actions.

Question 4

What is Natural Language Processing (NLP)?

Answer 4

NLP allows AI to process human languages, interpret context, and understand written and spoken language.

Question 5

What are the three main types of Machine Learning?

Answer 5

1. Supervised Learning (uses labeled data), 2. Unsupervised Learning (finds patterns without labels), 3. Reinforcement Learning (learns through trial and error in an environment).

Question 6

What are the two main types of Supervised Learning?

Answer 6

1. Regression - predicts continuous numeric values (e.g., temperature, house price), 2. Classification - predicts categories or classes (binary or multiclass).

Question 7

What is Deep Learning?

Answer 7

Deep learning is an advanced form and subset of machine learning that tries to emulate how the human brain learns, using neural networks with multiple layers (three or more define ‘deep’ learning).

Question 8

What are the main components of a Neural Network?

Answer 8

Neurons/Nodes (algorithms), Layers (input, hidden, output), Connections with Weights, Activation Functions, and Biases.

Question 9

What is Generative AI?

Answer 9

Generative AI is a subset of AI where models create original content in response to natural language prompts, including text, images, and programming code.

Question 10

What are Large Language Models (LLMs)?

Answer 10

LLMs are models trained on massive amounts of text data (billions/trillions of parameters) to understand and generate human-like language. Examples include GPT models.

Question 11

What are the key components of Transformer Architecture?

Answer 11

Tokenization (converts text to numeric tokens), Embeddings (semantic meaning vectors), Positional Encoding (word order), and Attention Layers (relationship quantification).

Question 12

What are Microsoft’s six Responsible AI principles?

Answer 12

1. Fairness, 2. Reliability and Safety, 3. Privacy and Security, 4. Inclusiveness, 5. Transparency, 6. Accountability.

Question 13

What are the three types of Azure AI resource types?

Answer 13

1. Single-service resources (specific to one AI capability), 2. Multi-service resources (bundle of services), 3. Azure OpenAI Service (access to OpenAI models).

Question 14

What is Azure Machine Learning?

Answer 14

A cloud service for managing the end-to-end ML lifecycle, including data preparation, model training (AutoML), evaluation, registration, and deployment through Azure Machine Learning Studio.

Question 15

What capabilities does Azure AI Vision provide?

Answer 15

Image analysis (captions, tags, object detection), smart cropping, background removal, facial detection and analysis, and Optical Character Recognition (OCR) using the Read API.

Question 16

What features does Azure AI Language offer?

Answer 16

Sentiment analysis, key phrase extraction, named entity recognition, language detection, summarization, question answering, and conversational language understanding (LUIS).

Question 17

What are the main capabilities of Azure AI Speech?

Answer 17

Speech-to-text (transcribing audio), text-to-speech (synthesizing audible speech), and speech translation capabilities.

Question 18

What is Azure AI Document Intelligence?

Answer 18

Formerly Forms Recognizer, it automatically reads, understands, and analyzes business documents (invoices, receipts, IDs, contracts), extracting structured data and key-value pairs using OCR.

Question 19

What is Azure AI Search used for?

Answer 19

Formerly Azure Cognitive Search, it enables knowledge mining by extracting insights from vast amounts of data, using indexers, skillsets, and creating searchable indexes.

Question 20

What is Azure AI Studio?

Answer 20

A unified web portal for AI development that brings together Azure Machine Learning, Azure OpenAI Service, and other Azure AI services into a single workspace for collaboration and development.

Question 21

What is Prompt Engineering?

Answer 21

The process of refining prompts (instructions) given to an AI application to generate higher quality and targeted responses, using techniques like system messages and clear expectations.

Question 22

What is Grounding in AI?

Answer 22

A prompt engineering technique where specific, relevant context is provided within a prompt to help the AI produce more accurate and related responses.

Question 23

What are Co-pilots in AI?

Answer 23

Generative AI assistants integrated into applications, often as chat interfaces, providing contextual support for common tasks (e.g., Microsoft Copilot, GitHub Copilot).

Question 24

What are the key components of Machine Learning training?

Answer 24

Training Data (features X and labels Y), Algorithm (finds relationships), Model (output function), and Inference (using the model for predictions).

Question 25

What is the difference between Small Language Models (SLMs) and Large Language Models (LLMs)?

Answer 25

SLMs are trained with smaller, subject-focused datasets and fewer parameters, effective for specific topics but less general. LLMs have massive parameters for broader capabilities but require more resources.

Question 26

What is Anomaly Detection in AI?

Answer 26

Anomaly Detection focuses on detecting outliers or suspicious patterns in data.

Question 27

What is Knowledge Mining?

Answer 27

Knowledge Mining involves extracting information and insights from large amounts of structured, semi-structured, or unstructured data, making it searchable and enriching it.

Question 28

What is Conversational AI?

Answer 28

Conversational AI involves technology that can participate in conversations with humans, such as chatbots and voice assistants.

Question 29

What is Clustering in Machine Learning?

Answer 29

Clustering is an unsupervised learning technique that groups observations based on similarities in their features (e.g., grouping flowers or customers).

Question 30

What are common evaluation metrics for Machine Learning models?

Answer 30

Mean Absolute Error (MAE) for regression models, and F1-score or Precision for classification models are used to assess model quality.

Question 31

What is the difference between single-service and multi-service Azure AI resources?

Answer 31

Single-service resources are specific to one AI capability with granular cost tracking and free tiers. Multi-service resources bundle several AI services with one endpoint and key but no free option.

Question 32

What makes Deep Learning training computationally intensive?

Answer 32

Deep learning models require significant compute power for iteratively passing training data through networks, evaluating predictions, and adjusting weights, often using GPUs for parallel operations.

Question 33

What is OCR and how is it used in Azure AI?

Answer 33

Optical Character Recognition (OCR) extracts text from images and documents. Azure uses the Read API to organize OCR results into pages, lines, and words.

Question 34

What is the purpose of validation data in Machine Learning?

Answer 34

Validation data is a portion of initial data set aside to test how close the model's predictions are to actual labels, helping assess model performance.

Question 35

What are the key AI workloads according to Microsoft?

Answer 35

Machine Learning, Computer Vision, Natural Language Processing, Conversational AI, Anomaly Detection, Knowledge Mining, and Generative AI.