Machine learning pioneer Michael I. Jordan dismisses AI doom rhetoric as grand moralizing performance

Science named Michael I. Jordan "the most influential computer scientist" in 2016. The UC Berkeley professor emeritus, now 70 and heading a new "markets and learning" chair at France's Inria in Paris, has given an interview to Libération in which he dismantles the past year's AI doom discourse as "a grand moralizing performance that doesn't serve the real objective it claims to pursue."

Jordan says he warned against the reckless overuse of the term "AI" back in 2019, and that he was not alone. "AI wasn't enough, so they started talking about AGI, and now it's superintelligence," he says — a term he calls science fiction. He dismisses AI companies' demands for regulation as theater: firms signal "look, we're the good guys" while wanting total control over how they are regulated.

His sharpest comments target the wave of resignations at AI labs, including a young Anthropic researcher whose early-September departure set off collective discussion of an "AIpocalypse." Jordan says many young employees at AI companies "aren't really engineers" — they come from the humanities and philosophy, are around 25, and have never worked where real products are built for real people. They gather, "a bit like in Parisian cafés in the 1920s," to discuss grand science-fiction ideas. When they resign, he says, they know they will attract attention, feel virtuous, and are presented as giving up money when they already hold their stock options.

He is especially critical of Nobel laureate Geoffrey Hinton, summarizing Hinton's argument — that we understand how the brain works, can reproduce it in silicon, copy it at will, and combine what the copies learn for exponential growth — and saying none of the steps is credible. "Geoffrey Hinton has never created a company that offers real services for real people." He is more generous toward Yoshua Bengio, whose warnings tend to target shortcomings in current technology, such as agent swarms spiraling out of control. The recent Hugging Face incident involving OpenAI agents, Jordan says, is simply bad engineering.

The interview lands as the doom economy matures. Safety institutes staff up on budgets that depend on the threat being taken seriously, a consulting industry grows around AI risk, and platform companies issue safety pledges that bind no one — Meta delayed its latest model for months over safety before releasing it anyway. Jordan's charge is that a consensus that cannot survive scrutiny from someone who has actually built the systems deserves exactly that scrutiny. He notes that "superintelligence" is now being pushed on the US tech elite by President Trump, who are happy to play along.

Jordan says the dominance of this discourse is harmful to technological development, to young people, and to the mindset of countries. The man Science called the field's most influential voice has not built a safety institute — he has built companies that offer real services for real people.