great piece by @lmatsakis Dario seems to have persuaded neither Washington nor Beijing that AI could pose grave risks. Part of the gap may be political and relate to some fundamental ways of understanding the world. Politicians and AI researchers have very different ideas about where power lies, and what could escape their control. But AI companies also have a communication problem. The OpenAI–Hugging Face incident is alarming once you understand what happened—but how can you explain to an auntie that this thing is actually dangerous and very different from computer worms??? My reading of “We Must Pace the Frontier” is that Dario is asking governments to help companies coordinate so safety work can keep up with AI development. That’s an enormous ask. If I were a politician without a technical background, I’d want a plain account of the what is the danger, the evidence, and exactly what I’m being asked to do—and why I should even listen. Frontier labs are trying to communicate a sense of urgency that governments don’t yet share, and I’ve realized the cognitive Venn diagram between Silicon Valley’s frontier AI companies and Washington is still very small. AI companies need to spend time translating urgent warnings into language that’s understandable and common sense. And so far “AI hacking Hugging Face is dangerous” and “AI will kill us all” are neither commonsensical nor understandable.