ryan_greenblatt comments on Counting arguments provide no evidence for AI doom

[ryan_greenblatt]

I think in Ajeya’s story the core threat model isn’t well described as scheming and is better described as seeking some proxy of reward.