One role subjective Bayesianism plays here is focusing discussions on forecasting the likelihood of AI risk rather than the conceptual possibility of AI risk. As someone who suspects “superintelligence” is basically conceptually unsound, this rings all kinds of alarm bells