Demis Hassabis has issued a strong warning about the rapid progress of artificial intelligence, saying that humanity may have only a few years to prepare for Artificial General Intelligence (AGI).
Speaking at Google DeepMind‘s annual developer conference, Hassabis said society is currently in the “foothills of the singularity,” a term often used to describe a future point when AI becomes as capable as, or even more capable than, humans in many tasks. He believes AGI could arrive within the next four years, reflecting growing confidence that researchers are on the right path to achieving this goal.
Hassabis explained that the next generation of AI agents should be viewed as a test run for much more advanced systems. According to him, these tools will show how governments, businesses, and society respond to increasingly powerful AI technologies. He pointed to recent advances in the industry as evidence that AI capabilities are improving faster than many people expected.
The Google DeepMind leader also stressed the need for governments and policymakers to take AI development seriously. He welcomed early efforts to introduce safety testing for advanced AI models but argued that stronger and faster action will be needed as technology continues to evolve.
Another area of concern is recursive self-improvement, where AI systems could eventually help improve future versions of themselves. While Hassabis said current systems cannot fully do this on their own, AI coding tools are already making engineers significantly more productive.
Addressing concerns about AI-related job losses, Hassabis argued that companies should use AI to increase innovation rather than reduce headcount. He believes higher productivity should allow organizations to pursue more ambitious projects, not simply cut jobs.
In fact, he said he would gladly hire talented engineers who have been laid off elsewhere, noting that there are countless opportunities across fields such as scientific research, healthcare, and game development.
