AI and safety · Question 31 of 60
Why does self-learning undermine the safety case of a system?
1 answer is correct. Medium
Show answer
Answer
- D Because the behavior progressively moves away from the originally tested behavior
Explanation
Before use, extensive testing shows that the system is safe. A self-learning system, however, keeps changing afterwards, and with every change the old evidence fits less well. Remedies are rules for how and from which data the model may learn, or built-in protective mechanisms.