Why can't we just disconnect a malevolent AI? Rob Miles on some of the simplistic solutions to AI safety.
