Shmi comments on AI risk, executive summary