Modeling distant superintelligences - History — LessWrong