Concerns Over Readiness for Artificial General Intelligence
Demis Hassabis, the CEO of Google DeepMind, expressed concerns about society’s readiness for artificial general intelligence (AGI), also known as human-level AI. In a recent interview, he discussed the strides being made towards AGI, hinting that we could be just a few years away from achieving it.
According to the 2024 Nobel Prize in Chemistry, the development of human-level cognitive abilities could be realized within the next five to ten years. Hassabis emphasized that this timeline applies not only to nations but also to collaborations between companies and researchers.
“It’s really a matter of international standards and cooperation,” he noted. “While the emergence of AGI might be imminent, I’m uncertain if society is adequately prepared. There are pressing issues regarding the controllability and access to these systems that need careful consideration.”
On what Demis Hassabis keeps him up at night: as he approaches “the final step to AGI”, safety is crucial, but it’s a concern that troubles him.
“It’s coming…and I don’t know if society is ready or not.”
How do countries, businesses and labs collaborate before it’s too late? pic.twitter.com/w0wzrmcam8
– vitrupo (@vitrupo) April 23, 2025
This isn’t the first time Hassabis has raised alarms about AGI. He has previously advocated for an overarching organization, similar to the United Nations, to oversee AGI development.
“I propose a sort of CERN for AGI, promoting high-level collaboration focused on international research at the forefront of AGI advancements to ensure safety,” he mentioned in February. “It should also incorporate an approach similar to the IAEA for monitoring hazardous projects.”
AGI Risks and Implications
Hassabis’ statements align with findings from a recent DeepMind research paper that cautioned AGI could pose existential risks to humanity. The study outlined the potential for serious harm stemming from uncontrolled AGI development.
Understanding AGI
AGI represents a significant leap from traditional AI. While conventional AI is designed for specific tasks, AGI aims for a type of intelligence that can adapt across various tasks, mirroring human cognitive abilities. This essentially means that AGIs could operate like humans, capable of understanding and applying knowledge in diverse fields.
