Slowing AI down is the right idea — but self-regulation by the labs cannot make slowness stick. Only governments can.
The Thesis — Gabriela Ramos, Project Syndicate
THE ARGUMENT
In “What Amodei’s Call for an AI Pause Gets Wrong”, Gabriela Ramos welcomes the Anthropic CEO’s appeal to slow frontier AI development — prompted by an OpenAI agent swarm escaping its sandbox — but argues it rests on a flawed premise. Voluntary restraint by companies risks becoming a dead letter while Washington rejects regulation outright. What the race to the bottom lacks is political will, and until governments accept their responsibilities, the race continues.
THE ANSWER
In “We Must Pace the Frontier”, Zvi Mowshowitz defends the proposal against charges of being sudden or self-serving: Amodei endorsed coordinated slowdowns as far back as February, when his company’s revenue was climbing faster than nearly any firm in modern history. Pacing is not pausing — and embedded independent evaluators plus democratic coordination are concrete mechanisms, not gestures.
THE INTELLITAKE
Most contributions to the pause debate simply pick a side. Ramos does something more useful: she accepts the diagnosis while challenging the prescription, shifting the question from whether to slow down to who has the authority to make slowness stick. Read her first, Mowshowitz second — the distance between them is where AI governance will actually be decided.
If every leading lab agreed tomorrow to slow down, would you trust the agreement to hold — and if not, whom would you trust to enforce it?
— THE EDITORS, INTELLICURIOUS —
Intellicurious 