Why I am cautious about deliberately slowing the development of advanced AI — and why safety should mean directing intelligence rather than capping it.