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

Michael I. Jordan, a machine learning pioneer whom Science described in 2016 as "the most influential computer scientist," has strongly criticized the current wave of warnings about existential AI threats. In an interview with the French newspaper Libération, the 70-year-old professor emeritus at UC Berkeley described the summer's AI discourse as "totally frustrating" and characterized much of the debate around superintelligence as belonging to the realm of science fiction. Jordan now works full-time at Inria, the French National Institute for Research in Digital Science and Technology in Paris, where he heads the newly created "markets and learning" chair.
Jordan said he had warned about the reckless overuse of the term AI back in 2019. He recalled that the escalation was predictable: the label AI eventually stopped being enough, so the field moved to AGI, and now to superintelligence. In his view, occasional discussion of far-fetched scenarios behind the scenes is acceptable, but when such thinking becomes the dominant discourse it is "very, very harmful" — for technological development, for young people and for the mindset of countries.
A central target of his criticism is the way some AI companies have positioned themselves on regulation. Jordan called their demands for regulation "theater," describing it as a way of signaling that they are the good guys while seeking to retain control over how they are regulated.
He was equally dismissive of the resignations of young AI researchers. He cited the departure of Jacob Coxon, a young Anthropic researcher, which in early September triggered a wave of collective anxiety about an "AIpocalypse" scenario. Jordan said many young employees at such companies are not really engineers — he described them as coming from the humanities, philosophy or other disciplines, around 25 years old, naive, not doing engineering work and never having worked at large companies where real products are developed for real people with discipline. He compared their gatherings to discussions in Parisian cafés in the 1920s about grand science-fiction ideas. When they resign, he said, they know they will attract attention; it lets them feel virtuous and be seen as important figures. He added that they claim to be giving up money even though they already hold their stock options. Jordan summed this up as a grand moralizing performance that does not serve the objective it claims to pursue, saying it combines naiveté with a certain cynicism.
On superintelligence itself, Jordan was blunt. He called it a science-fiction term and said a large part of the debates he has witnessed belong to the realm of science fiction. He was especially critical of Nobel laureate Geoffrey Hinton, saying Hinton has never created a company that offers real services for real people. Jordan summarized Hinton's argument as follows: we have figured out how the brain works, we can reproduce it in silicon, we can copy it as many times as we like, and by combining what all the copies learn we would obtain exponential growth leading to superintelligence. Jordan's verdict on that chain of reasoning was that none of the steps is credible and that it is science fiction.
He was somewhat more generous toward Yoshua Bengio, another recipient of the Turing Award alongside Hinton. Jordan said Bengio is a little different because he takes action and does things that focus directly on safety. Bengio's warnings, according to Jordan, tend to target shortcomings in current AI technology rather than speculative superintelligence, giving the example of agent swarms spiraling out of control.
Jordan also addressed the hacking incident involving Hugging Face and OpenAI agents. He framed it as bad engineering rather than an existential signal. In chemical engineering, he said, we would never have allowed explosions to happen just to see how badly things could go wrong. He described OpenAI's work as conducting rather stupid experiments, letting agents do whatever they wanted and encouraging that behavior. Doomers, he said, then loved turning it into something scary — the idea that the agents are self-aware and trying to destroy humans — which he also considers science fiction.
On the economics of the field, Jordan views the enormous investments in data centers as just an enormous bet with no certainty that it will pay off. Current large language models are already capable enough for natural language tasks, and trillions spent on data centers will not necessarily deliver much more, in his view. He argued that what is often needed instead is a bit of specialized knowledge in a particular field based on local knowledge, and that this does not require a gigantic data center, particularly if open-weight models are integrated into a company.
Why it matters: For developers and product teams, Jordan's critique is a reminder that the loudest narratives around AI risk do not necessarily reflect engineering consensus, and that decisions about model size and compute should be tied to concrete tasks rather than speculative timelines. His emphasis on smaller, specialized, open-weight models integrated locally suggests a practical alternative to chaining every product roadmap to ever-larger data-center buildouts. His comments also sharpen a debate that shapes hiring, research funding and regulation, though whether his characterization of younger researchers or of corporate regulatory positions is fair remains contested — that is interpretation, not established fact.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Michael I. Jordan, described by Science as "the most influential computer scientist," considers warnings about superintelligence pure science fiction. He calls the resignations of young AI researchers a "grand moralizing performance" and companies' demands for regulation "theater." On Hinton's thesis that the brain can be reproduced in silicon and scaled into superintelligence, he's blunt: "None of the steps in this argument is credible. Here, we're in the land of science fiction." The article Machine learning pion