Logan Zoellner comments on Predictable updating about AI risk