AI Challenges & Responsibility

Question 1

A model makes decisions about who can obtain loans. Bias in mind, which core dimension of Responsible AI is relevant?

Accepted Answer

Explainability

Question 2

Security & Privacy for AI systems include:

Accepted Answer

Secure the cloud computing platform, edge devices, data stores, Encryption of data at rest and in transit, Conduct security assessment, penetration testing and code reviews, Deploy AI based threat detection systems

Question 3

Where can you find information about the responsible AI practices that AWS has implemented for their AI services?

Accepted Answer

AWS AI Service Cards

Question 4

A Gen AI chat outputs sensitive PII data from its training data into responses. What risk is this illustrating?

Accepted Answer

Exposure

Question 5

Sagemaker Clarify is...

Accepted Answer

Full model evaluation on accuracy, robustness, toxicity, Bias detection

Question 6

Some core dimensions of responsible AI include....

Accepted Answer

Fairness, Explainability, Privacy & Security

Question 7

Which of the following is NOT a challenge associated with responsible AI?

Accepted Answer

Scalability

Question 8

I need to consult responsible AI docs to understand the AWS AI service, its features, use cases & limitations.

Accepted Answer

Use AI Service Cards

Question 9

Partial dependence plots are useful because...

Accepted Answer

They depict how a single feature can influence the predicted outcome

Question 10

What isn’t a capability of Gen AI?

Accepted Answer

Determinism

Question 11

What are challenges of generative AI?

Accepted Answer

Data security, privacy concerns, nondeterminism, regulatory violations, interpretability, Social risks, toxicity, hallucinations, plagiarism / cheating

Question 12

The purpose of MLOps is to ensure models aren't just developed but also

Accepted Answer

deployed, monitored, retrained systematically & repeatedly

Question 13

How would you define interpretability for a model?

Accepted Answer

Transparency into a system so a human can explain the model’s output

Question 14

In the AWS Shared Responsibility Model, what is AWS responsible for?

Accepted Answer

Securing the cloud infrastructure, including hardware, software, networking

Question 15

Which Human Centered Design principle should one use to help decision makers prevent mistakes in high-pressure envs?

Accepted Answer

Design for amplified decision-making

Question 16

What is Responsible AI?

Accepted Answer

Mitigating risks & negative outcomes that can emanate from an AI system

Question 17

The relationship between AI interpretability & AI performance can be best described as...

Accepted Answer

Inverse relationship

Question 18

What pipeline stage is not apart of MLOps?

Accepted Answer

Monitoring & deployment, Data and Building & Testing

Question 19

Amazon Bedrock is...

Accepted Answer

Human or automatic model evaluation

Question 20

What are the key principles for MLOps?

Accepted Answer

Continuous retraining, continuous monitering, continuous delivery, Version control, automation, continuous integretion

Question 21

The importance of Governance & Compliance in AI includes:

Accepted Answer

Managing, optimizing, and scaling the organizational AI initiative, Clear policies & guidelines to ensure legal & regulatory requirements

Question 22

What are the five GenAI scopes you can classify your application as?

Accepted Answer

Consumer, enterprise, pre-trained, fine-tuned, and self-trained models

Question 23

Security Generative AI demands Governance & Compliance, Risk management, Controls, Resilience, and Legal & Privacy

Accepted Answer

True

Question 24

As per the Human Centered Design approach for explainable AI, we should design for...

Accepted Answer

amplified decision making, human and AI learning, unbiased decision making

Question 25

Which of the following helps with processes that define, implement, and enforce compliance?

Accepted Answer

Governance

AI Challenges & Responsibility

Questions 25

AI Challenges & Responsibility

Questions 25