A non-logarithmic argument for Kelly — AI Alignment Forum