Q1. Scenario: A stakeholder asks: "What's the difference between AI, ML, and Deep Learning? Use an analogy with building a house."
AI is the goal of building a house (intelligent behavior). ML is using blueprints from data (learning patterns). Deep Learning is using specialized machinery (neural networks) to build very complex structures, e.g., multi-floor designs.
Q2. Scenario: You are asked to recommend a solution for predicting stock prices. Would you choose traditional programming, ML, or DL? Why?
ML (e.g., linear regression, random forests) because stock data has patterns but not extremely high-dimensional. DL could overfit due to limited data. Traditional rules are impossible to hand-code.
Q3. Scenario: A client wants to detect faces in images. Explain why deep learning is preferred over classical ML (e.g., SVM with HOG features).
Deep learning automatically learns hierarchical features from raw pixels; classical ML requires handcrafted features (e.g., HOG). Deep learning scales with data and achieves higher accuracy on complex visual tasks.
Q4. Scenario: You have a dataset of 100 customers to predict churn. Which approach would you use: rule-based, ML, or DL? Why?
Rule-based if patterns are simple and known. ML (e.g., logistic regression) with small data is acceptable. DL would overfit severely. Rule-based may be too brittle; ML is a balance.
Q5. Scenario: A chess engine uses minimax search without learning. Is this AI? Is it ML? Why?
It is AI (problem-solving by search), but not ML (no learning from data). It uses predefined heuristics. ML would involve learning evaluation functions from game outcomes.