Reasoning through arguments against taking AI safety seriously - Yoshua Bengio