Sublime

Aligning machine goals with our own involves three unsolved problems: making machines learn them, adopt them and retain them.

Studies have shown that while such results-based approaches can deliver progress in the short term, in the long term they risk collapse. Two academics from Oxford studying the National Health Service in the UK recently remarked:

By tapping preexisting intuitions and translating information into mind-friendly formats, it’s possible to hone people’s statistical reasoning. Hone we must. Risk literacy is essential for doctors, judges, policymakers, and others who hold our lives in their hands. And since we all live in a world in which God plays dice, fluency in Bayesian

In his important book After Virtue, Alasdair MacIntyre famously says, “I cannot answer the question, ‘What ought I to do?’ unless I first answer the question, ‘Of which story am I a part?’”

and present danger to the future of liberty.

So how could social labs help avert this apocalypse? There are three strategic responses based on the ideas presented in this book—stabilization, mitigation, and adaptation. Again, these are not silver bullets, but they demonstrate that practical responses are well within our means.

People are not monads. A real, live human self is always already partial to certain, select others. Morality needs to take this essential fact about human selfhood into account rather than pretend to override it.

Two prominent philosophers proved to be especially influential members. The first was Michael Sandel, a Harvard professor who is the contemporary successor to John Rawls in defining the concept of justice. At the time, he was writing an essay titled “The Case Against Perfection: What’s Wrong with Designer Children, Bionic Athletes, and Genetic