Question
What are the core principles of responsible AI?
Click to reveal answer
Answer
Fairness, transparency, explainability, accountability, privacy, safety, and human oversight.
Click to flip back
All Responsible AI Flashcards
Q: What are the core principles of responsible AI?
A: Fairness, transparency, explainability, accountability, privacy, safety, and human oversight.
Q: What is data bias?
A: When training data does not accurately represent the target population, leading to skewed model predictions that unfairly favor or disadvantage certain groups.
Q: What is a model card?
A: A document that describes a model's intended use, training data, performance metrics, limitations, and ethical considerations.
Q: What is explainability in AI?
A: The ability to understand and interpret why an AI system made a specific decision or prediction.
Q: What is human-in-the-loop?
A: A design pattern where human reviewers validate or correct AI outputs before they are acted upon, especially for high-stakes decisions.
Q: What AWS service detects bias in ML models?
A: Amazon SageMaker Clarify — it detects bias in training data and model predictions, and provides feature importance for explainability.
Q: How does Amazon Bedrock Guardrails prevent harmful content?
A: Through configurable content filters (hate, insults, sexual, violence), denied topics, word filters, sensitive information (PII) filters, and contextual grounding.
Q: What is prompt injection?
A: A security attack where malicious input attempts to override the system prompt or manipulate the model into producing unintended outputs.
Q: What is Amazon A2I?
A: Amazon Augmented AI — a service for building human review workflows to review ML predictions, especially for sensitive or uncertain outputs.
Q: What is model drift?
A: When a model's performance degrades over time because the real-world data distribution changes compared to training data. Detected by SageMaker Model Monitor.