Question 1

A bank is developing its model risk governance framework to address ML/AI models. What foundational approach should the bank take?

Options :

A : Create an entirely new framework specific to AI/ML.

B : Build upon its existing model risk framework while adapting to new challenges from AI/ML.

C : Exclude traditional models from the risk framework.

D : Focus solely on AI/ML ethical aspects without considering statistical issues.

Answer: B

Question 2

A retail bank uses self-training to classify loan applicants as high or low risk, but it finds that updating the model after each new labeled data point is computationally intensive. Which approach can the bank use to reduce this burden?

Options :

A : Select multiple best-predicted instances in each round

B : Use an unsupervised learning model instead

C : Increase the complexity of the initial model

D : Use transductive learning instead

Answer: A

Question 3

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 4

An admissions committee decides to incorporate a measure of resilience alongside grades in its AI-based selection criteria. Which challenge is the committee likely to face with this approach to individual fairnes

Options :

A : Defining resilience as a measurable trait

B : Avoiding direct discrimination

C : Ensuring grades remain the sole basis for admissions

D : Guaranteeing equal admissions for all applicants

Answer: A

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