Rohin Shah comments on [AN #59] How arguments for AI risk have changed over time