mattmacdermott comments on Bengio’s Alignment Proposal: “Towards a Cautious Scientist AI with Convergent Safety Bounds”