✅ Generative AI – Solved MCQs (50 Questions) 🔹 Section 1: Basics of Generative AI 1. What is the primary goal of generative AI ? A. Da...
✅ Generative AI – Solved MCQs (50 Questions)
🔹 Section 1: Basics of Generative AI
1. What is the primary goal of generative AI?
A. Data compression
B. Classification
C. Creating new data that mimics original data
D. Data deletion
✅ Correct Answer: C. Creating new data that mimics original data
2. Which of the following is NOT a type of generative model?
A. GANs
B. Autoencoders
C. CNNs
D. Diffusion Models
✅ Correct Answer: C. CNNs
3. GAN stands for:
A. General AI Network
B. Generative Adversarial Network
C. Generic Artificial Neural
D. Guided AI Network
✅ Correct Answer: B. Generative Adversarial Network
4. Who introduced GANs in 2014?
A. Geoffrey Hinton
B. Ian Goodfellow
C. Yann LeCun
D. Andrew Ng
✅ Correct Answer: B. Ian Goodfellow
5. A typical GAN consists of:
A. Encoder and Decoder
B. Generator and Discriminator
C. Classifier and Filter
D. Reader and Writer
✅ Correct Answer: B. Generator and Discriminator
6. Variational Autoencoders (VAEs) are used for:
A. Image classification
B. Feature extraction
C. Data generation through probabilistic encoding
D. Reinforcement learning
✅ Correct Answer: C. Data generation through probabilistic encoding
7. Which model is best known for generating text?
A. GPT
B. YOLO
C. ResNet
D. LeNet
✅ Correct Answer: A. GPT
8. Which of the following is a diffusion-based generative model?
A. BERT
B. StyleGAN
C. DALL·E 2
D. ResNet
✅ Correct Answer: C. DALL·E 2
9. What type of data can generative AI create?
A. Images
B. Music
C. Text
D. All of the above
✅ Correct Answer: D. All of the above
10. The discriminator in GANs tries to:
A. Generate realistic samples
B. Fool the generator
C. Distinguish real from fake data
D. Encode data
✅ Correct Answer: C. Distinguish real from fake data
🔹 Section 2: Applications of Generative AI
11. Which of the following is NOT a use case of generative AI?
A. Image enhancement
B. Speech synthesis
C. Data anonymization
D. Sorting numbers
✅ Correct Answer: D. Sorting numbers
12. Deepfake technology is based on:
A. Clustering
B. Regression
C. Generative models
D. Decision trees
✅ Correct Answer: C. Generative models
13. Which company developed ChatGPT?
A. Google
B. Meta
C. OpenAI
D. DeepMind
✅ Correct Answer: C. OpenAI
14. DALL·E is capable of:
A. Image captioning
B. Translating languages
C. Text-to-image generation
D. Video compression
✅ Correct Answer: C. Text-to-image generation
15. Which model is typically used for text generation?
A. GAN
B. LSTM
C. GPT
D. ResNet
✅ Correct Answer: C. GPT
16. Music generation can be done using:
A. GANs
B. RNNs
C. Transformers
D. All of the above
✅ Correct Answer: D. All of the above
17. Which tool generates AI art from text prompts?
A. TensorFlow
B. Midjourney
C. PyTorch
D. Scikit-learn
✅ Correct Answer: B. Midjourney
18. A generative AI model that writes code is:
A. Copilot
B. YOLOv8
C. VGG
D. RetinaNet
✅ Correct Answer: A. Copilot
19. Generative AI in drug discovery is used to:
A. Scan barcodes
B. Generate molecule structures
C. Track infections
D. Prescribe dosage
✅ Correct Answer: B. Generate molecule structures
20. A business use of generative AI includes:
A. Handwriting detection
B. Writing marketing content
C. License plate recognition
D. Spam filtering
✅ Correct Answer: B. Writing marketing content
🔹 Section 3: Techniques & Architectures
21. Transformers are primarily used in:
A. Regression
B. Text generation and translation
C. Object detection
D. Clustering
✅ Correct Answer: B. Text generation and translation
22. GPT is an acronym for:
A. General Prediction Tree
B. Generative Pre-trained Transformer
C. Gradient Prediction Transformer
D. Google Pre-trained Tokenizer
✅ Correct Answer: B. Generative Pre-trained Transformer
23. StyleGAN is mainly used for:
A. Detecting spam
B. Speech recognition
C. Human face generation
D. Neural search
✅ Correct Answer: C. Human face generation
24. What loss function is common in GANs?
A. Cross-entropy loss
B. Hinge loss
C. MSE
D. Triplet loss
✅ Correct Answer: A. Cross-entropy loss
25. The encoder in a VAE compresses:
A. Data to a fixed label
B. Input into a latent distribution
C. Images into pixels
D. None of the above
✅ Correct Answer: B. Input into a latent distribution
26. What is the major limitation of early GANs?
A. Slow training
B. Mode collapse
C. Low accuracy
D. Inflexible data
✅ Correct Answer: B. Mode collapse
27. In diffusion models, the generation process involves:
A. Blurring the input
B. Adding and removing noise
C. Gradient descent
D. Decision trees
✅ Correct Answer: B. Adding and removing noise
28. Autoencoders consist of:
A. Two GANs
B. Encoder and Decoder
C. Tokenizer and Classifier
D. Generator and Filter
✅ Correct Answer: B. Encoder and Decoder
29. Transformers learn using:
A. LSTM memory
B. Attention mechanism
C. Filters
D. Backpropagation only
✅ Correct Answer: B. Attention mechanism
30. Conditional GANs (cGANs) generate data based on:
A. Noise alone
B. Image size
C. Conditioned input (e.g., labels)
D. Loss rate
✅ Correct Answer: C. Conditioned input (e.g., labels)
🔹 Section 4: Ethical & Societal Aspects
31. A major ethical concern in generative AI is:
A. Faster speed
B. Resource saving
C. Misinformation and deepfakes
D. File size
✅ Correct Answer: C. Misinformation and deepfakes
32. Which law applies to AI-generated copyrighted content?
A. Copyright Act
B. Digital Media Bill
C. No universal law yet
D. IT Act
✅ Correct Answer: C. No universal law yet
33. AI-generated content should be labeled to ensure:
A. Code reuse
B. Attribution
C. Transparency and trust
D. Better fonts
✅ Correct Answer: C. Transparency and trust
34. Generative models trained on biased data can:
A. Improve ethics
B. Produce accurate results
C. Amplify societal bias
D. Save cost
✅ Correct Answer: C. Amplify societal bias
35. Deepfake videos threaten:
A. Gaming
B. Security and public trust
C. Signal processing
D. Traffic systems
✅ Correct Answer: B. Security and public trust
36. Over-reliance on generative AI can:
A. Improve writing skills
B. Reduce human creativity
C. Create accurate maps
D. Increase reading speed
✅ Correct Answer: B. Reduce human creativity
37. Generative AI raises concerns in:
A. Supply chain
B. Facial recognition
C. Data privacy and authorship
D. Sports
✅ Correct Answer: C. Data privacy and authorship
38. Watermarking AI-generated content helps in:
A. Style transfer
B. Compression
C. Authenticity detection
D. Translation
✅ Correct Answer: C. Authenticity detection
39. Generative AI models can hallucinate, meaning:
A. They think
B. They visualize real data
C. They produce factually incorrect but confident outputs
D. They dream
✅ Correct Answer: C. They produce factually incorrect but confident outputs
40. Mitigating bias in generative AI requires:
A. Faster GPUs
B. Data balancing and transparency
C. Random noise injection
D. More images
✅ Correct Answer: B. Data balancing and transparency
🔹 Section 5: Advanced Concepts
41. Latent space in generative models refers to:
A. A storage unit
B. The compressed representation of data
C. A training algorithm
D. GPU space
✅ Correct Answer: B. The compressed representation of data
42. Prompt engineering is key in:
A. CNNs
B. Traditional ML
C. Text-to-image and text-to-text models
D. Filtering spam
✅ Correct Answer: C. Text-to-image and text-to-text models
43. The temperature parameter in text generation controls:
A. Token size
B. Creativity/randomness of output
C. Memory
D. Compression ratio
✅ Correct Answer: B. Creativity/randomness of output
44. Diffusion models outperform GANs in:
A. Overfitting
B. Image realism and diversity
C. Video compression
D. Speed
✅ Correct Answer: B. Image realism and diversity
45. A typical transformer model uses:
A. RNNs
B. Convolutions
C. Self-attention layers
D. PCA
✅ Correct Answer: C. Self-attention layers
46. LLMs like GPT are pre-trained on:
A. Medical data only
B. Video datasets
C. Large text corpora
D. GPS signals
✅ Correct Answer: C. Large text corpora
47. Fine-tuning refers to:
A. Training from scratch
B. Adjusting hyperparameters
C. Updating a pre-trained model on a specific task
D. Removing layers
✅ Correct Answer: C. Updating a pre-trained model on a specific task
48. The main input for a text-to-image model is:
A. A sketch
B. An image
C. A sentence or text prompt
D. A color palette
✅ Correct Answer: C. A sentence or text prompt
49. Which generative model was developed by Stability AI?
A. BERT
B. DALL·E
C. Stable Diffusion
D. T5
✅ Correct Answer: C. Stable Diffusion
50. One way to evaluate generative models is through:
A. Accuracy
B. FID (Fréchet Inception Distance)
C. ROC curve
D. AUC
✅ Correct Answer: B. FID (Fréchet Inception Distance)
COMMENTS