Michael I. Jordan calls AI doom talk "grand moralizing performance"

Jordan's critique is not new in kind — he warned about the reckless overuse of the term AI back in 2019 — but he now frames the escalated vocabulary as a predictable progression. "AI wasn't enough, so they started talking about AGI, and now it's 'superintelligence'," he says. That framing positions the doom discourse as a rebranding exercise rather than a technical finding, which is why he treats the summer's talk as "totally frustrating".

The resignations at Anthropic matter to Jordan as a case study in incentives rather than evidence. He points out that the departing researchers "already have their stock options" and argues that resigning itself guarantees attention — a mechanism that, in his account, rewards visibility over engineering judgment.

He draws the same line on research safety. He describes the Hugging Face hack by OpenAI agents as bad engineering, comparing it to allowing explosions in chemical engineering just to see how badly things could go wrong, and says the doomers loved turning it into something scary. Against the trillion-dollar data center build-out he is similarly skeptical, calling it "an enormous bet, but we have absolutely no idea whether it will pay off," and suggesting open-weight models integrated into a company may serve specialized needs without a gigantic data center. Jordan also singles out Yoshua Bengio as different from other prominent warners, saying Bengio takes action focused directly on safety and points to shortcomings such as agent swarms spiraling out of control.