Question 1

An insurance firm uses a predictive model to detect fraud. In the interest of balancing sensitivity and specificity, which of the following challenges might the firm face?

Options :

A : Maximizing sensitivity without increasing the false-positive rate

B : Reducing the base rate to ensure fair outcomes

C : Ensuring predictive rate parity and equal opportunity simultaneously

D : Applying demographic parity to improve prediction quality

Answer: A

Question 2

A generative AI provider in China is designing its service to align with Chinese regulations. What is a key obligation it must meet under the Generative AI Measures?

Options :

A : Ensure all content generated is free from any factual errors.

B : Prioritize content generation for government-endorsed topics only.

C : Prevent the generation of illegal and discriminatory content, respecting privacy rights.

D : Avoid transparency in algorithm usage to maintain competitive advantage.

Answer: C

Question 3

A bank is using a multinomial logit model to predict customers' choice among different loan products: personal loan, home loan, or credit card. If personal loan is set as the baseline category, which statement correctly describes the probability of choosing a credit card?

Options :

A : The probability is the same as the probability of not choosing a home loan

B : It must be calculated as a separate probability from the baseline

C : It is one minus the probability of choosing a personal loan

D : It is calculated as the product of probabilities for each category

Answer: B

Question 4

A critical infrastructure provider in the U.S. wants to ensure responsible AI deployment to enhance service reliability. How should the provider proceed under guidance from the AI Safety and Security Board (AISSB)?

Options :

A : Only focus on AI deployment in non-critical areas to minimize risk.

B : Deploy AI solutions that solely increase service efficiency without security considerations.

C : Limit AI usage to what is already industry-standard to avoid compliance issues.

D : Implement recommendations from the AISSB to address AI-related disruptions to critical services, protecting national security and public health.

Answer: D

Question 5

A financial institution uses a neural network model with thousands of parameters to predict loan defaults. On the training dataset, the model has a nearly zero residual sum of squares (RSS). However, it performs poorly on new data. What does this indicate?

Options :

A : The model has insufficient parameters

B : The model is underfitted

C : The model is overfitted

D : The model has perfect generalization

Answer: C