Hassabis frames AGI as near‑term civilisational shift

Demis Hassabis is talking about artificial general intelligence less like the next chatbot upgrade and more like a machine that could reorder whole industries within a few years. In an April 2026 conversation, the Google DeepMind chief said it could land in about five years and hit with the force of “ten industrial revolutions compressed into a single decade.” (thetwentyminutevc.com, the-decoder.com) He was not describing today’s tools as finished. In the same interview, Hassabis said artificial intelligence is “overhyped in the short term” but “underhyped in the long term,” which is his way of saying current products are noisy while the deeper shift is still ahead. (singjupost.com, the-decoder.com) Artificial general intelligence means a system that can handle many kinds of cognitive work instead of one narrow task. Hassabis has repeatedly framed it as software that can learn, plan, reason, and adapt across domains, closer to a general-purpose worker than a single-purpose calculator. (cnbc.com, thetwentyminutevc.com) That helps explain why he keeps reaching for Industrial Revolution language. The first Industrial Revolution changed textiles, transport, factories, and cities over many decades, and Hassabis is arguing that general intelligence in software could push through science, medicine, robotics, and office work on a much tighter clock. (britannica.com, the-decoder.com) Google DeepMind’s own project list shows why he talks that way. The lab is still building chat models, but its flagship science work includes AlphaFold, which maps the three-dimensional shapes of proteins, and Gemini Robotics, which is built to help robots perceive, reason, and act in the physical world. (deepmind.google, deepmind.google) AlphaFold is the clearest example of the pitch. Google DeepMind says the system has revealed millions of protein structures, and its newer AlphaFold 3 model was presented as a tool for predicting how proteins, DNA, ribonucleic acid, and drug-like molecules interact. (deepmind.google, blog.google) The drug story is no longer just a research demo. Isomorphic Labs, the Alphabet-linked company built around this biology push, raised $600 million on March 31, 2025 to advance its artificial-intelligence drug design engine and move therapeutic programs toward the clinic. (isomorphiclabs.com) The robotics story points the same way. In March 2025, Google DeepMind introduced Gemini Robotics as a Gemini 2.0-based model meant to bring multimodal reasoning into machines that can manipulate objects, use tools, and respond in real-world spaces. (deepmind.google, deepmind.google) So when Hassabis warns about a civilisational shift, he is not only talking about better search boxes or faster coding assistants. He is describing a world where the same core intelligence stack spills from screens into laboratories, warehouses, and drug pipelines. (thetwentyminutevc.com, deepmind.google) That framing also matches how the lab sells itself to recruits. Google DeepMind’s careers page says its research scientists need “deep theoretical knowledge” and “an interdisciplinary mindset,” which is a very different hiring message from a company chasing only the next quarterly product launch. (deepmind.google) Hassabis has been giving versions of this timeline for more than a year, but the April 2026 version is sharper because it pairs urgency with restraint. Five years is close enough to shape hiring, spending, and policy now, while the warning about short-term hype gives DeepMind room to keep backing slow, expensive science bets that may pay off after the current product cycle. (cnbc.com, singjupost.com, thetwentyminutevc.com)