7.7. Practice Questions: Generative AI

Test your understanding of generative AI concepts, Azure OpenAI, and responsible AI for GenAI.

Domain 5: Generative AI (Questions 41-50)

Question 41: Which model generates images from natural language prompts?

• A. GPT-4
• B. DALL-E ✓
• C. Embeddings
• D. Whisper

Rationale: DALL-E generates images from text descriptions. GPT generates text, Embeddings create vectors, Whisper transcribes speech.

Question 42: Which capability is NOT supported by DALL-E?

• A. New image creation
• B. Creating image variations
• C. Editing images
• D. Image description ✓

Rationale: DALL-E generates images but cannot describe or analyze them. Image description requires different models.

Question 43: [Answer choice] can return responses such as natural language, images, or code based on natural language input.

• A. Computer vision
• B. Machine learning
• C. Generative AI ✓
• D. OCR

Rationale: Generative AI models can generate various content types (text, images, code) from natural language prompts.

Question 44: Which layer applies content filters in the responsible generative AI model?

• A. Model layer
• B. Metaprompt and grounding
• C. Safety system ✓
• D. User experience

Rationale: The safety system layer includes content filters that classify and suppress harmful content.

Question 45: [Answer choice] can identify constraints and styles for generative AI responses.

• A. Data grounding
• B. System messages ✓
• C. Content filters
• D. Embeddings

Rationale: System messages set context by describing expectations, constraints, and behavioral guidelines for the model.

Question 46: What is the FIRST stage when developing a responsible generative AI solution per NIST?

• A. Build content filters
• B. Identify potential harms ✓
• C. Deploy the model
• D. Train the model

Rationale: According to NIST AI Risk Management Framework, identifying potential harms is the first stage.

Question 47: [Answer choice] uses plugins to help users with tasks from a generative AI model.

• A. Embeddings
• B. DALL-E
• C. Copilots ✓
• D. Transformers

Rationale: Copilots are AI assistants integrated into applications that use plugins to extend capabilities.

Question 48: Which TWO are GPT model capabilities?

• A. Create natural language ✓
• B. Detect specific dialects
• C. Understand natural language ✓
• D. Generate physical products

Rationale: GPT models excel at understanding and creating natural language. They don't detect dialects or create physical products.

Question 49: [Answer choice] can search, classify, and compare text for similarity.

• A. Data grounding
• B. System messages
• C. Embeddings ✓
• D. DALL-E

Rationale: Embeddings convert text to numerical vectors, enabling similarity search, classification, and comparison.

Question 50: Which THREE are image generation features of generative AI models?

• A. Animation of static images
• B. Creating variations of an image ✓
• C. Editing an image ✓
• D. New image creation ✓

Rationale: Generative AI can create new images, variations, and edits. Animation is not a standard image generation feature.