Arturs comments on Predictable updating about AI risk