Hassabis and the 'foothills of singularity': what he said and what it means

At the close of the Google I/O keynote on May 20, 2026, Demis Hassabis, CEO of Google DeepMind, made a statement now circulating across every industry forum: we are in the "foothills of the singularity". He framed it within a speech about how Google's research and products will help unlock AGI's potential "for the benefit of all humanity". The Verge covered the moment and, wisely, added a question mark to the headline.

This is not the first time a top executive in the industry has reached for eschatological language to describe AI's current state. But coming from Hassabis, a researcher with solid scientific credentials rather than a corporate communicator, it warrants more careful analysis than the usual headline cycle.

What he said exactly

According to The Verge's coverage, Hassabis claimed that when we look back at this moment, we will understand we were in the "foothills" of something much larger. He did not say singularity is already here, nor did he name a specific date. The geographical metaphor is deliberate: we are climbing, but the main peak is still not visible.

The term "technological singularity" has a long and contested history. Popularized by Vernor Vinge in 1993 and amplified by Ray Kurzweil, it describes the hypothetical point at which artificial intelligence surpasses human cognitive capacity in a sustained and autonomous way, triggering change that in theory escapes any prediction. That Hassabis uses it at a marketing stage like Google I/O is not trivial, but it is also not a formalized technical prediction.

Why it matters beyond the rhetoric

Two possible readings exist for this statement, and neither is dismissible.

The first is pragmatic: Hassabis is competing in a market where OpenAI and Anthropic also use high-voltage language to attract talent, investment, and regulatory attention. Speaking of "foothills of singularity" at the close of a keynote is, in that context, a positioning move. Google DeepMind needs to project ambition at a moment when its models, Gemini included, have faced criticism for trailing behind competitors on certain benchmarks.

The second reading is more technical and, if Hassabis sustains it in academic settings, deserves to be taken seriously: the capability jumps observed over the last 18 months in multimodal reasoning, long-term planning, and use of external tools are qualitatively different from those in previous cycles. Not because the models are "conscious", something no serious person claims, but because the curve of improvement in tasks once considered exclusively human has changed its slope.

For whom this debate is relevant

For engineering teams working with models like Claude Opus 4.7 or with the full Gemini suite in production, statements like this have an indirect but real impact: when CEOs of the major labs speak of imminent AGI, investment cycles accelerate, ethics committees activate, and most importantly for daily work, provider roadmaps reorient toward autonomous reasoning capabilities and long-duration agents.

For product leads and executives making decisions about AI adoption, the risk is different: confusing keynote enthusiasm with reliable technical roadmap. The gap between "we are in the foothills" and "this will change how my company operates in six months" is not guaranteed by any Google I/O announcement.

For researchers and academics, the most honest question is whether Hassabis has internal data about systems Google DeepMind is developing in the lab that justifies that claim, or whether he is projecting an optimistic narrative about trends everyone already observes in public benchmarks. That distinction matters, and for now there is no way to verify it from outside.

Editorial perspective

That Demis Hassabis uses the word "singularity" at a keynote does not make it a scientific prediction, but it does not empty it of content either: it comes from someone who knows exactly what they are saying and to whom. It would be worth monitoring closely whether this language migrates from marketing stages to DeepMind papers.