How can we look at the error rates of current AI (especially the ones dealing with human aspects) and subsequently think that this is the right time to make an

How can we look at the error rates of current AI (especially the ones dealing with human aspects) and subsequently think that this is the right time to make an AI-controlled weapon? Absolutely stunning. I guess it doesn't matter what the error rate is when you think everyone else on the planet is your enemy.