ryan_greenblatt comments on How do we solve the alignment problem?

[ryan_greenblatt]

Should be possible for agents with long run preferences to strategy steal, so I don’t see why evolution is an issue from this perspective.