I don't understand why anyone would think low-time-preference as counter-intuitive. It's perfectly intuitive. But addressing language models like they have a personality does.