Infinite Data/Compute Arguments in Alignment — LessWrong