Michael I. Jordan rejects AI doom claims

Michael I. Jordan, professor emeritus at UC Berkeley and now working full-time at Inria in Paris, gave a wide-ranging interview to Libération about the current AI debate. He said the discussion around existential risk had become frustrating and insisted that much of it belonged to science fiction.

Jordan said the language of AI, AGI and superintelligence had expanded in ways that encouraged hysteria. He argued that companies calling for regulation were performing virtue while seeking to control how they were regulated.

He pointed to the resignation of Anthropic researcher Jacob Coxon as an example of what he saw as performative behaviour by younger staff. In his account, some of them came from non-engineering backgrounds, wanted attention, and treated resignation as a moral gesture.

Jordan was more pointed in his criticism of Geoffrey Hinton, saying the case for copying the brain in silicon and combining the results did not hold together. He said Yoshua Bengio was different because his work focused directly on safety, and he also described OpenAI's agent experiments and the resulting hack of Hugging Face as bad engineering rather than evidence of self-aware systems.

On spending, Jordan said the rush into data centres was only a large bet and that nobody knew whether it would pay off. He argued that many practical uses of language models would need specialised knowledge rather than ever-larger infrastructure.