AI-900 Question 55
Select 3An organization is implementing an AI-powered medical diagnosis system. Which considerations are essential to ensure the system is reliable and safe for use in this critical application?
- A
Ensure the training data is diverse and representative of the population it will serve.
- B
Regularly monitor the AI system for biases and update it as needed.
- C
Prioritize maximizing performance metrics like accuracy over all other considerations.
- D
Incorporate human oversight to validate critical decisions made by the AI system.
- E
Disable logging and auditing features to protect patient privacy.
Show answer and explanation
Correct answers: A, B, D
Explanation
AI solutions in critical applications like medical diagnosis require careful consideration of reliability and safety. This involves using diverse and representative training data to avoid biases, continuously monitoring and updating the system, and incorporating human oversight to validate decisions. While accuracy is important, it should not come at the expense of fairness or safety. Additionally, transparency through logging and auditing is essential to maintain accountability and improve system reliability.
- A. Correct.
Ensuring the training data is diverse and representative is critical to avoid biases and ensure fair and reliable outcomes for all user groups. Using non-representative data could lead to unsafe decisions in a medical context.
- B. Correct.
Monitoring for biases and updating the system as needed ensures that the AI system remains reliable and fair over time, especially as new data or scenarios emerge.
- C. Incorrect.
While performance metrics like accuracy are important, prioritizing them over other considerations such as fairness, safety, and reliability can lead to harmful outcomes, especially in critical scenarios like medical diagnosis.
- D. Correct.
Human oversight ensures that decisions with high stakes, such as medical diagnoses, can be validated and corrected if necessary, enhancing the safety and reliability of the AI system.
- E. Incorrect.
Disabling logging and auditing features can compromise transparency and prevent effective troubleshooting or accountability, which are essential for ensuring safety and reliability.