Comment by madmulita

How would a malfunctioning AI decide it's safe to stop?