Shmi comments on Three AI Safety Related Ideas