Beyond the Turing Test

There is rampant speculation by people, including Bostrum, who have never written any Machine Learning code that Hard AI will spontaneously appear. Anyone who has implemented ML knows that this is silly given the heavy manual effort of feature selection and data cleansing, coupled with the relative simplicity of today's approaches based on gradients and stats.

I'm glad the author also raises the point of machine ethics. Here again, humans are anthropomorphising our ethics. Human ethics evolved from living in a cooperating community (See Krebs, "Origins of Morality"). Machine ethics may be quite different and hopefully significantly more strategic than human selfishness which has resulted in horrible damage to other species, our host earth and of course millions of daily suffering people.