Michael I. Jordan, a pioneer of machine learning, rebuts AI doom rhetoric as a grand moralizing performance

Meta description: Michael I. Jordan calls AI doom talk science fiction, criticizing Anthropic resignations, Geoffrey Hinton, OpenAI agents, and data center trillions.

Tags: AI doom, Michael I. Jordan, Anthropic, OpenAI, data centers

Featured image: Abstract neural-network lines over server racks and market graphs, no people, no logos, no text.

A Long-Standing Warning Against AI Hype

Michael I. Jordan, whom Science called the most influential computer scientist in 2016, has spent years resisting inflated language around artificial intelligence. In a wide-ranging interview with Libération, the 70-year-old machine learning pioneer called AI doom talk pure science fiction and said he takes aim at doomers, company founders, and naive junior researchers. He described the summer's AI discourse as totally frustrating.

Jordan is professor emeritus at UC Berkeley and now works full-time at Inria, the French National Institute for Research in Digital Science and Technology in Paris, where he heads the new "markets and learning" chair. His criticism is not new. He said he warned about the reckless overuse of the term AI in 2019, noting that he was not alone and that the resulting hysteria was obvious. In his account, the escalation followed a clear path: AI was not enough, so the talk moved to artificial general intelligence, and then to superintelligence.

Resignations, Regulation, and Moralizing Performance

Jordan dismissed AI companies' demands for regulation as theater, a way of saying they are the good guys while they would like total control over how they are regulated.

The resignation of Jacob Coxon, a young Anthropic researcher, became an example for him after it sparked a wave of collective hysteria over an AIpocalypse scenario in early September. Jordan said many young employees at these companies are not really engineers. He described them as people from humanities, philosophy, or other disciplines, often 25 years old, naive, and without experience building real products for real people inside large companies with discipline. He compared their discussions to Parisian cafés in the 1920s, where grand science-fiction ideas were debated.

According to Jordan, resignations can attract attention, allow people to feel virtuous, and present them as important figures. He said they may claim to be giving up money, but stock options are often already in place. For him, this is a grand moralizing performance that does not serve the objective it claims to pursue, combining naivety with a certain cynicism.

Superintelligence, Hinton, Bengio, and the Hugging Face Episode

Jordan treats superintelligence as a science-fiction term and says much of the debate belongs to that realm. He does not object to discussing far-fetched scenarios behind the scenes, but he warned that dominant discourse becomes very harmful for technological development, young people, and the mindset of countries. He noted that President Trump is pushing the term on the US tech elite, who are happy to play along.

His sharpest criticism is aimed at Nobel laureate Geoffrey Hinton. Jordan said Hinton has a track record of bold claims that do not always hold up and has never created a company offering real services for real people. Hinton's argument, in Jordan's retelling, assumes the brain can be understood and reproduced in silicon, copied many times, and combined to produce exponential growth that becomes superintelligence. Jordan said none of the steps are credible.

He is slightly more positive about Yoshua Bengio. Jordan said Bengio differs by taking action and doing things that focus directly on safety. He added that Bengio's warnings tend to target shortcomings in current AI technology, such as agent swarms spiraling out of control.

Jordan also addressed the hack of Hugging Face by OpenAI agents. He called it bad engineering, using a chemical-engineering analogy: no one would allow explosions just to see how badly things could go wrong. He said OpenAI carried out rather stupid experiments by letting agents do whatever they wanted and encouraging that behavior. Doomers, he argued, then turned the episode into a scary story about self-aware agents trying to destroy humans, which Jordan again dismissed as science fiction.

The Trillion-Dollar Data Center Question

Jordan considers the trillion-dollar bet on data centers just an enormous bet, adding that he has absolutely no idea whether it will pay off. LLMs, he said, are already capable enough for natural language tasks, and trillions spent on data centers will not necessarily deliver much more.

Instead, Jordan emphasized adding a bit of specialized knowledge in a particular field based on local knowledge. He said a gigantic data center is not needed for that, particularly when companies use open-weight models and integrate them internally. The interview closes with a practical question: whether AI progress will continue to be shaped by speculative warnings, or by smaller, grounded uses of existing models and tools.