RobertM comments on OpenAI: Our approach to AI safety