How do we train AI to not make the same mistakes that humans do?

How do we train AI to not make the same mistakes that humans do?