6.2. Common Exam Patterns and Distractors

Pattern 1: Workload Confusion

• Questions mix up similar workloads
• Know: OCR ≠ NLP (OCR is Computer Vision)
• Know: Object Detection ≠ Image Classification (bounding boxes = detection)

Pattern 2: Algorithm Type Confusion

• Logistic Regression sounds like regression but is CLASSIFICATION
• Clustering is UNSUPERVISED; regression/classification are SUPERVISED
• Multiple linear regression = 2+ features; linear regression = 1 feature

Pattern 3: Responsible AI Principle Confusion

• Fairness = bias, discrimination, demographic factors
• Accountability = legal standards, compliance, governance
• Transparency = users understanding limitations

Pattern 4: Service Capability Confusion

• Translator = text-to-text ONLY
• DALL-E = generates images, does NOT describe them
• GPT = text generation, NOT dialect detection

Pattern 5: "Which TWO" Questions

• Often ask for two related concepts
• Both answers must be correct—partial credit rare
• Eliminate answers that don't fit together