At Google's flagship I/O developer conference, Google DeepMind CEO Demis Hassabis outlined why he believes the singularity—the theoretical point where AI surpasses human intelligence and begins self-improving—is closer than ever. In an interview with Axios cofounder Mike Allen, Hassabis attributed this acceleration directly to the rise of powerful AI agents capable of autonomously building things for humans.
As a concrete example of this new reality, Hassabis shared that he has been using AI to build mini video games late at night—creative and technical tasks that historically would have taken teams of developers many months to complete. "This year, with the agentic systems that we're all seeing and using, I think we can start feeling it now," he remarked, emphasizing the palpable shift in AI capabilities.
Furthermore, Hassabis projected that Artificial General Intelligence (AGI), where machines reach human-level intelligence across broad domains, will arrive as soon as 2030. He argued that the societal and economic impact of AI is still vastly underestimated, declaring that it will ultimately be 100 times more impactful than the Industrial Revolution. However, the DeepMind chief dismissed sci-fi-style doomsday scenarios of machines taking over the world.
[AgentUpdate Depth Analysis] Hassabis’s emphasis on "agentic systems" as the primary catalyst for the singularity marks a fundamental shift in the AI paradigm. We are transitioning from static, conversational LLMs to dynamic, goal-oriented AI Agents characterized by reasoning, planning, and tool integration. Hassabis’s personal anecdote of generating video games overnight highlights a democratization of software engineering where natural language becomes the ultimate programming interface. Compared to previous orchestrations, modern agent frameworks (such as AutoGen, LangGraph, and Google's own agentic initiatives) are closing the loop of self-improvement. For the broader AI Agent ecosystem, this signals that the race is no longer just about raw model parameters, but about building robust multi-agent systems that can autonomously solve complex, multi-step workflows. This evolutionary loop is the true engine that will drive us toward AGI by 2030.
