On "The contagion of fear"

The comparison between personal software mistakes and existential AI risk collapses under any scrutiny, and using it to dismiss legitimate researcher concern is exactly the kind of false equivalence that weakens the credibility of anyone pushing back on AI doomism. Cantrill's story might feel reassuring—mistakes happen, panic subsides, the world moves on—but researchers at Anthropic have access to actual model internals and scaling trends; they're not panicking based on youthful blunders. If you're trying to argue that AI safety concerns are overblown, you need a better counterargument than "I once broke something and people freaked out unnecessarily." This kind of rhetorical move just signals you don't have one.