The AI Safety Panic Has No Clothes

When Google DeepMind safety researcher Bilal Choktai resigned last month and declared that AI "has the potential to kill us all," he joined a growing chorus of insiders issuing apocalyptic warnings. Jacob Coxon's viral resignation from Anthropic made similar claims. At least eleven AI safety researchers have now publicly demanded a slowdown. But as Ed Zitron, host of Better Offline and author of the Where's Your Ed At newsletter, points out to host Ed Elson on Prof G Markets, none of them have explained what they're actually afraid of — or what "slowing down" would even mean.

The Safety Movement's Hollow Core

Zitron traces the origins of AI doomerism to the rationalist and effective altruist (EA) communities that have heavily populated AI labs. He describes them as operating like a religious cult: "They believe it on very flimsy terms. They don't really have any evidence for it other than going, 'Look, a thing we vaguely predicted kind of worked.'" The core claim — that AI will undergo "recursive self-improvement" and rapidly become uncontrollable — rests on theoretical projections, not demonstrated capabilities.