MS Champ Month of Jan

What is Responsible AI?

• a) An approach to developing AI safely and ethically

• b) A new machine-learning algorithm

• c) A type of neural network architecture

• d) A dashboard for analyzing models

Which of these is not one of the six principles of Microsoft's Responsible AI Standard? 

• a) Fairness

• b) Reliability & Safety

• c) Accuracy

• d) Inclusiveness

What does the Responsible AI Dashboard provide? 

• a) Access Azure Open AI Services

• b) Automated model training 

• c) Data labelling

• d) A central place to access Responsible AI tools

What are some components of the Responsible AI Dashboard?

• a) Model debugging and business decision-making tools

• b) Data visualization and database management 

• c) Automated data labelling and augmentation

• d) Neural architecture search and hyperparameter tuning

How can you find model performance inconsistencies? 

• a) Compare metrics across cohorts

• b) Analyze probability distributions

• c) Visualize performance scores  

• d) All of these

What are some reasons to use the Responsible AI Dashboard?  

• a) Centralize access to Responsible AI tools 

• b) Streamline workflows

• c) Improve tool integration

• d) All of these

What do most of the traditional ML metrics focus on?  

• a) Ratio of correct vs incorrect predictions

• b) Training time

• c) Loss function values

• d) Number of parameters

What metrics can show model performance for a cohort? 

• a) Accuracy, Precision, Recall, MAE, RMSE

• b) Learning rate, batch size, epochs 

• c) Loss, gradient norm, step size

• d) Flops, latency, throughput

What can probability distributions show? 

• a) Uncertainty estimates

• b) The distribution of a cohort's predicted outcomes

• c) Decision Boundaries

• d) Feature importance

Model explainability can help with:  

• a) Compliance and auditing

• b) Identifying biases

• c) Increasing transparency

• d) All of these