Rather than asking, "What's your p(doom)?" we should be asking, "Under what conditions does AI risk increase or decrease?" ...