Google DeepMind CEO Demis Hassabis has declared that artificial general intelligence could reshape human civilization more profoundly than the discovery of fire or the harnessing of electricity, and he believes the technology is only a handful of years away.
A Civilizational Turning Point
Hassabis, who leads Google's flagship AI research lab, framed AGI as a transformation of unprecedented scale, placing it above the most pivotal breakthroughs in human history. In his view, a machine capable of matching or exceeding human cognitive abilities across virtually any task would fundamentally alter how societies function, produce, and solve problems.
The DeepMind chief has consistently argued that AGI is not a distant science-fiction concept but a near-term reality. By his timeline, the arrival of systems with broad, human-level intelligence could occur within just a few years rather than decades.
AGI won't just be another tool — it could redefine what human civilization is capable of achieving.
That optimism about capability, however, comes paired with a call for caution. Hassabis has repeatedly emphasized that the immense power of such systems demands careful, coordinated oversight before they are unleashed on the world.
A Push for Independent Testing
To address the risks, Hassabis proposed the creation of a dedicated U.S. standards body tasked with evaluating frontier AI models before they reach the public. The idea is to establish a rigorous, independent checkpoint that vets the most advanced systems for safety and reliability.
Such an institution would function similarly to regulatory bodies in other high-stakes industries, giving governments and developers a shared framework for assessing whether powerful models are ready for release.
Key elements of the proposal include:
- An independent authority to test frontier AI systems
- Safety evaluations conducted before models are deployed
- A standardized approach that developers across the industry could follow
The suggestion arrives amid intensifying debate over how quickly AI capabilities are advancing and whether existing oversight mechanisms can keep pace. As competition among leading labs accelerates, Hassabis's comments underscore a growing tension between racing to build transformative technology and ensuring it can be safely controlled.