private_messaging comments on How can we ensure that a Friendly AI team will be sane enough?