Demis Hassabis, CEO of Google DeepMind and described in the original piece as a Nobel laureate, published an essay this week claiming that AGI could emerge within a few years.

What Hassabis defines as AGI

He distinguishes ordinary AI systems from AGI, portraying the latter as a system capable of human-like cognition across science, creativity, strategy and self-directed learning.

Hassabis compares the significance of AGI not to incremental technologies but to transformative discoveries such as fire or electricity in historical consequence.

Risks and regulatory proposal

The essay warns that current models already strain cybersecurity and could enable biological or nuclear threats beyond typical software vulnerabilities.

To address these risks, Hassabis proposes creating a U.S. regulatory body modeled on FINRA, tasked with testing frontier models prior to public release.

Industry response and noted irony

The proposal received endorsements via reposts from Sam Altman, Jack Clark, Mustafa Suleyman, Satya Nadella and Sundar Pichai, among others.

The piece notes an irony: many of these figures, including Sam Altman, Elon Musk, Ilya Sutskever and Dario Amodei, were founding members of OpenAI in 2015, originally opposing a dominant Google-led AGI.

Ten years later, Hassabis calls for unified oversight while former critics publicly acknowledge the need for regulation.

Concentration, stakes and social contrast

Commentary highlights the geographic concentration of the AI race in parts of San Francisco and the long-term investments by Google in research since the early 2000s.

The text emphasizes that delays of even a week can translate into billions of dollars and, potentially, real human costs under current competitive dynamics.

It also contrasts the scale of technological developments with the day-to-day priorities of most people, and questions how far model capabilities may advance under ongoing leverage by leading AI companies.