Question 1

In the context of AI ethics, what is the "black box" problem, and why is it a concern?

Options :

A : The black box problem is the challenge of integrating AI systems with existing legacy systems in an organization.

B : The black box problem occurs when AI models are too complex, leading to inefficient training and inference times.

C : The black box problem is the lack of transparency and interpretability in AI models, making it difficult to understand how decisions are made.

D : The black box problem refers to the difficulty in accessing and using AI technologies due to proprietary constraints.

Answer: C

Question 2

Why might one choose a linear kernel over a non-linear kernel in SVMs?

Options :

A : Linear kernels provide better scalability in high dimensions.

B : Linear kernels are more efficient for large, sparse datasets.

C : Non-linear kernels require more support vectors to perform well.

D : Non-linear kernels tend to overfit small training datasets.

Answer: B

Question 3

What method would you use to apply a ReLU activation function element-wise in a PyTorch model?

Options :

A : torch.nn.functional.relu()

B : torch.nn.ReLU()

C : apply_relu()

D : relu()

Answer: A

Question 4

Which approach is most effective for feature selection in large-scale data?

Options :

A : LASSO

B : Random Forest

C : Recursive Feature Elimination

D : PCA

Answer: A

Question 5

Which type of residual pattern suggests a poorly fitting linear model?

Options :

A : Random scatter plot

B : Uniform linear trend

C : Funnel shape pattern

D : Vertical clustering

Answer: C