Google Professional Machine Learning Engineer Question 458
Select 3Google Cloud PlatformYou are designing an AI solution for a financial institution to automate loan approvals. During the development process, you must assess potential risks associated with the AI model. Which of the following considerations should you include to identify and mitigate risks to the AI solution?
- A
Bias in training data leading to unfair decisions for certain demographics.
- B
Model explainability to ensure stakeholders understand how decisions are made.
- C
Cloud storage costs for storing training datasets.
- D
Security and privacy of sensitive customer data used for training.
- E
Versioning of machine learning models for reproducibility.
Show answer and explanation
Correct answers: A, B, D
Explanation
When identifying risks to AI solutions, it is crucial to consider issues like bias, explainability, and the security of sensitive data, as these directly impact the ethical, regulatory, and operational integrity of the system. While other factors like storage costs and model versioning are important, they are not directly related to risk identification.
- A. Correct.
Bias in training data is a critical risk in AI solutions, especially in sensitive domains like finance. If not addressed, it can lead to unfair decisions and regulatory repercussions.
- B. Correct.
Model explainability is essential to ensure transparency and trust in the AI system, particularly when decisions impact customers directly.
- C. Incorrect.
While cloud storage costs are relevant for budgeting, they are not directly linked to the risks posed by the AI solution itself.
- D. Correct.
The security and privacy of sensitive customer data are key considerations to prevent data breaches and maintain compliance with regulations.
- E. Incorrect.
Versioning of machine learning models is important for tracking changes, but it is not specifically tied to identifying risks to the AI solution.