JanB comments on Mitigating extreme AI risks amid rapid progress [Linkpost]