22 AI experts and scholars "Preparing for an 'intelligence explosion' where AI creates AI... could lead to human extinction"

Major AI researchers have warned that rapid automation of AI development could lead to an "intelligence explosion" where technological advancements happen at an unprecedented pace, potentially leading to humans losing control over highly advanced AI systems. In a joint paper, 22 researchers from top institutions, including Turing award winners Geoffrey Hinton, Yoshua Bengio, and Andrew Barto, noted that AI development is increasingly being automated, with some companies like Anthropic seeing an 80% increase in AI-generated code.

The researchers estimate that AI research and development projects could be fully automated by mid-2028, leading to a significant acceleration in AI advancements. They are urging governments to take action to mitigate the risks of an intelligence explosion, including establishing independent third-party evaluators to assess AI systems and monitor AI development within companies.