BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 4 — Questions and Answers
Question 1: What distinguishes a generative AI model from a discriminative model?
- Generative models only classify inputs; discriminative models also produce outputs
- Generative models learn P(X|Y) or P(X) to synthesize data; discriminative models learn P(Y|X) for classification (Correct answer)
- Generative models require labeled data; discriminative models do not
- Generative models use rule-based systems; discriminative models use neural networks
Correct answer: Generative models learn P(X|Y) or P(X) to synthesize data; discriminative models learn P(Y|X) for classification
Generative models learn data distributions to generate new samples, while discriminative models learn boundaries to classify existing inputs.
Question 2: In computer vision, what is the purpose of a convolutional layer in a CNN?
- To flatten spatial features into a 1D vector for classification
- To apply learned filters across the input to detect local spatial patterns (Correct answer)
- To normalize activations across a batch of inputs
- To pool adjacent pixels by taking their maximum value
Correct answer: To apply learned filters across the input to detect local spatial patterns
Convolutional layers slide learned filters over the input, computing dot products to detect features like edges, textures, or shapes.
Question 3: What is the Frame Problem in AI?
- Difficulty in representing images as numerical frames for processing
- The challenge of specifying what does NOT change when an action is performed in a world model (Correct answer)
- Limitations of decision trees when feature count grows exponentially
- Memory overflow when storing all possible game states
Correct answer: The challenge of specifying what does NOT change when an action is performed in a world model
The Frame Problem is the challenge of efficiently representing and reasoning about which aspects of the world remain unchanged after an action.
Question 4: Which algorithm is used in Monte Carlo Tree Search (MCTS) to balance exploration and exploitation of the game tree?
- Minimax with alpha-beta pruning
- Upper Confidence Bound applied to Trees (UCT) (Correct answer)
- Expectimax algorithm
- Iterative deepening A*
Correct answer: Upper Confidence Bound applied to Trees (UCT)
UCT selects nodes using UCB1 formula balancing exploitation (high-value nodes) and exploration (less-visited nodes).
Question 5: In transfer learning, what is 'fine-tuning'?
- Training a model from scratch on a target dataset
- Adjusting pre-trained model weights on a new task-specific dataset with a low learning rate (Correct answer)
- Replacing the final layer of a model and freezing all other layers permanently
- Randomly reinitializing weights to adapt to a new domain
Correct answer: Adjusting pre-trained model weights on a new task-specific dataset with a low learning rate
Fine-tuning continues training a pre-trained model on new data, typically with a small learning rate to preserve learned representations.
Question 6: What is the primary function of the attention mechanism in transformer models?
- To apply convolutions over sequential input tokens
- To weight the relevance of different input tokens when producing each output representation (Correct answer)
- To encode positional information using sine and cosine functions
- To normalize hidden states across the sequence length
Correct answer: To weight the relevance of different input tokens when producing each output representation
Attention computes a weighted sum of value vectors, where weights reflect how relevant each key-value pair is to the current query.
Question 7: Which ethical principle in AI refers to the ability to understand and explain how an AI system reached a specific decision?
- Fairness
- Robustness
- Explainability / Interpretability (Correct answer)
- Privacy
Correct answer: Explainability / Interpretability
Explainability (or interpretability) ensures that AI decisions can be understood and justified by humans, which is critical for trust and accountability.
BSCS Bachelor of Science in Computer Science: Artificial Intelligence Fundamentals 4
What distinguishes a generative AI model from a discriminative model?