We are animals that make tools. That is what defines us.
Demis Hassabis, the Nobel laureate who leads Google DeepMind, has offered a measured reframing of artificial intelligence — not as a harbinger of apocalypse, but as the latest in humanity's long lineage of tools built to extend perception and understanding. Speaking with The Economist, he acknowledged that serious risks exist if the technology is not developed with care, while insisting that what separates catastrophe from progress is not fate but deliberate, collective effort. His is the voice of someone who helped build something powerful and now carries the weight of knowing that capability and wisdom do not always arrive together.
- AI is advancing faster than the ethical and safety frameworks designed to govern it, creating a gap that Hassabis openly acknowledges could lead to serious harm.
- The temptation to frame AI as either savior or destroyer distorts the conversation — Hassabis pushes back against both extremes, insisting the technology is a tool, not a deity.
- Concrete promises — curing disease, solving the energy crisis — sit uneasily alongside concrete warnings, forcing the field to hold hope and caution in the same hand.
- No single company or nation can navigate this alone; Hassabis calls for the world's best minds to collaborate across borders on the problem of safe development.
- The trajectory is neither assured nor doomed — it lands, for now, on a conditional optimism: humanity can get this right, but only if it chooses to treat the stakes seriously.
Demis Hassabis occupies a rare position — a neuroscientist and computer scientist who runs Google DeepMind, won the 2024 Nobel Prize in Chemistry for AI-driven protein folding research, and now speaks with unusual authority about both what artificial intelligence can do and what it might cost us.
In a recent conversation with The Economist, he offered a deliberate reframing: AI is not a god, not a monster, but a tool — sophisticated beyond anything we have built before, yet still in the lineage of the telescope and the microscope. Those instruments extended human perception without replacing human judgment. Hassabis sees AI the same way, as the latest expression of our species' defining capacity to build things that let us see further into reality.
The potential he describes is grounded, not abstract. Protein folding research is already reshaping drug discovery. Energy and disease remain solvable problems if the tools are applied well. But Hassabis does not look away from the other side. He acknowledged plainly that things could go quite wrong if the technology is not built correctly — the words of someone who understands that good intentions and good outcomes are not the same thing.
What he believes can close that gap is not luck but sustained, serious work — the best minds across companies and nations collaborating on safety, treating the problem with the gravity it deserves. His confidence in human ingenuity is real, but it is conditional. We can get this right, he suggests. Getting it right, however, will require more than belief.
Demis Hassabis sits at the intersection of two worlds that rarely meet cleanly: the laboratory and the boardroom. The British neuroscientist and computer scientist runs Google DeepMind, the research division that has pushed artificial intelligence further than most thought possible. In 2024, he won the Nobel Prize in Chemistry for his work on computational protein design—using AI to predict how proteins fold, a breakthrough that could reshape medicine. He is, by any measure, one of the people who understands what AI can do.
But understanding capability and understanding risk are not the same thing. In a recent conversation with The Economist, Hassabis offered a careful reframing of how we should think about artificial intelligence altogether. He rejected the framing that treats AI as something godlike or apocalyptic. Instead, he positioned it as what humans have always made: a tool. A very sophisticated one, yes. But a tool nonetheless.
"I think of this as a tool," he said. "That's how we're trying to build it: as a scientific instrument, like a telescope or a microscope." The comparison matters. Telescopes and microscopes extended human perception. They didn't replace human judgment; they amplified it. They let us see further, deeper, into the structure of things. Hassabis sees AI in that lineage—as the latest in a long chain of instruments humans have built to interrogate reality. What defines us as a species, he suggested, is precisely this capacity to make tools. We built civilization with our general intelligence and our hands. Now we're building something that might let us see further still.
The potential is real. Hassabis pointed to concrete possibilities: curing diseases, solving the energy crisis. These are not speculative benefits. Protein folding alone has already begun to reshape drug discovery. The tools work. But he did not minimize the other side of the ledger. "There is a possibility that things could go quite wrong if the technology is not built in the right way," he acknowledged. This is not the language of someone who sleeps easily. It is the language of someone who knows what he has helped create and understands that intention and outcome do not always align.
What separates catastrophe from success, in his view, is not luck. It is time, care, and concentration. It is the best minds in the world—not just at DeepMind, but across the industry, across nations—working together on the problem of how to build this safely. He expressed confidence in human ingenuity, but it was a conditional confidence. We can get this right, he seemed to say, if we treat it with the seriousness it deserves.
For Hassabis, the people drawn to this work are driven by something fundamental: an insatiable curiosity and a hunger for knowledge. They are following an old tradition, just with newer tools. The question now is whether that tradition—the one that has given us so much—can be stewarded carefully enough to avoid the pitfalls that come with power. He believes it can. But belief, he made clear, is not enough. It requires work.
Notable Quotes
I think of this as a tool. That's how we're trying to build it: as a scientific instrument, like a telescope or a microscope.— Demis Hassabis
There is a possibility that things could go quite wrong if the technology is not built in the right way. I am very optimistic if we have enough time, care and concentration.— Demis Hassabis
The Hearth Conversation Another angle on the story
When you say AI is a tool like a telescope, aren't you downplaying something? A telescope doesn't learn. It doesn't improve itself.
That's fair. But the point isn't that they're identical. It's that they serve the same function: they extend what we can know. A telescope let us see Jupiter's moons. AI lets us see patterns in protein sequences we couldn't see before. Both are instruments of discovery.
But you also said things could go "quite wrong." What does that actually mean? What are you worried about?
I think he's being careful not to name specific doomsday scenarios, but he's acknowledging that power without wisdom is dangerous. If you build something very capable and don't think through the consequences, you could cause harm. That's not speculation—it's history.
So the solution is just... more smart people working on it?
Not just more people. The right people, working together across companies and countries, with time and focus. He's saying this can't be rushed. It can't be a race. It has to be deliberate.
Do you think he actually believes we'll get it right?
I think he believes we can. But I also think he knows that belief alone doesn't guarantee anything. That's why he keeps saying it requires care and concentration. He's not confident because he's optimistic. He's optimistic because he thinks we're capable of being careful.