They link to Dario's blog, but his first "critical" point is to

>[avoid] thinking about AI risks in a quasi-religious way