Eric Schmidt Says America Won't Pause AI — Why That One Word Matters

Eric Schmidt, the former CEO of Google, went on CNN and said something blunt: America is not going to pause AI development, because the entire incentive system pushes every company to keep going, and a ban couldn't be enforced across all the labs anyway. That sounds reasonable. But an AI-governance writer on the forum LessWrong says the framing hides a trick — and the trick is the word "pause" itself.

Here's the argument, in plain terms. We already accept that companies don't build things that would obviously destroy them. Nobody needs a law forcing a carmaker not to sell exploding cars; it just isn't done. If everyone genuinely agreed that a certain AI was that dangerous, we'd say labs "can't" build it, not that they're bravely "pausing." Calling it a pause makes ordinary caution sound like a strange, radical, economy-wrecking move that needs special justification. That framing hands the burden of proof to the cautious side, and lets the racing side off the hook.

The writer also flags a real practical wrinkle: when competitors publicly agree to slow down together, that can look like illegal price-fixing. So a voluntary "pause" invites lawsuits, while a shared, matter-of-fact understanding that something is too risky to build does not. Words, in other words, have legal and market consequences.

A commenter pushed back fairly: if you claim everyone secretly knows what "the dumb thing" is, you should be able to name it — was releasing GPT-4 the mistake? Everything after? The author concedes he's pushing a frame, but argues his frame is better and more honest. Whatever you decide, one thing is clear: how we name the choice shapes whether slowing down looks sane or strange.