Podcast
Questions and Answers
When would a prediction be sent for human review?
When would a prediction be sent for human review?
- When the model has low confidence in the prediction. (correct)
- When the client application requests it.
- When the model has high confidence in the prediction.
- When the prediction is flagged for potential bias.
What is the purpose of Amazon Augmented AI?
What is the purpose of Amazon Augmented AI?
- To automate the process of building machine learning models from scratch.
- To create custom machine learning models tailored to specific business needs.
- To improve the accuracy of machine learning models by incorporating human feedback. (correct)
- To provide a platform for users to collaborate on machine learning projects.
What is the primary source of human reviewers for Amazon Augmented AI?
What is the primary source of human reviewers for Amazon Augmented AI?
- AWS employees only.
- AWS contractors only.
- AWS Mechanical Turk.
- A combination of AWS employees, contractors, and Mechanical Turk workers. (correct)
What happens to the human-reviewed predictions in Amazon Augmented AI?
What happens to the human-reviewed predictions in Amazon Augmented AI?
What is the role of the 'risk-weighted scores' in Amazon Augmented AI?
What is the role of the 'risk-weighted scores' in Amazon Augmented AI?
Flashcards
Amazon Augmented AI (A2I)
Amazon Augmented AI (A2I)
A service that combines machine learning predictions with human oversight.
Human Oversight
Human Oversight
Involvement of humans in reviewing machine learning predictions to ensure accuracy.
Confidence Score
Confidence Score
A metric indicating how certain a machine learning model is about its prediction.
AWS Mechanical Turk
AWS Mechanical Turk
Signup and view all the flashcards
Risk-Weighted Scores
Risk-Weighted Scores
Signup and view all the flashcards
Study Notes
Amazon Augmented AI (A2I)
- A2I enables human oversight of machine learning models in production.
- Input data is fed into an AWS AI service or a custom machine learning model, which produces predictions.
- High-confidence predictions are returned immediately to the client application.
- Low-confidence predictions are sent to human reviewers.
- Human reviewers consolidate predictions and assign risk-weighted scores stored in Amazon History.
- Client applications receive predictions, including those reviewed by humans.
- Reviewed predictions are fed back into the machine learning model to improve accuracy.
Human Reviewers in A2I
- Reviewers can be company employees, AWS contractors (over 500,000), AWS Mechanical Turk users, or pre-screened vendors.
- A wide range of reviewers guarantees adequate confidentiality.
Model Deployment for A2I
- Machine learning models can be based on AWS services (e.g., Amazon Rekognition).
- Models can be built using Amazon SageMaker.
- Models can be hosted externally, integrated with A2I.
Studying That Suits You
Use AI to generate personalized quizzes and flashcards to suit your learning preferences.
