The Man Who Wants to Cure Cancer Thinks We're Still Underestimating This

Demis Hassabis built AlphaFold. Now he's saying the 10-year picture is still too small.

Demis Hassabis spent his career building toward AGI. He founded DeepMind, solved protein folding, and spun out Isomorphic Labs to attack drug discovery. When he says something is underestimated, it's worth paying attention.

His view: AI is overhyped in the next year. And still massively underestimated over 10.

Both things are true simultaneously. The short-term noise is real — the demos, the breathless announcements, the things that don't quite work yet. But the 10-year picture is a different category of change. He puts it this way: AGI is like 10 Industrial Revolutions at 10 times the speed. A century of upheaval compressed into a decade.

What He's Actually Building

AlphaFold solved protein structure. That was the foundation. Now Isomorphic Labs is working on the full drug design pipeline — designing compounds, checking toxicity, validating properties. His estimate: five to ten years to have that engine ready.

The bottleneck after that isn't science. It's regulatory. Clinical trials still take a decade. His path forward: get a dozen AI-designed drugs through the full process, let regulators back-test the model predictions against real outcomes, and eventually — maybe ten years beyond that — compress the process by trusting the models enough to skip some steps.

The interviewer's mother has MS. Hassabis didn't offer false comfort. He offered a timeline that's honest and still more optimistic than conventional medicine allows itself to be.

The Question Nobody Is Asking

He's been thinking about something that isn't in the economic or technical debates.

Assume we get the safety right. Assume we solve the inequality problem. Then what?

What is meaning when intelligence is no longer scarce? What is purpose when the things that previously required human effort — creativity, discovery, problem-solving — can be done faster and better by machines? What does it mean to be human in a world where cognition is abundant?

He doesn't have the answers. He thinks we need new philosophers, not just new engineers.

This is the question the next decade will force — not whether AI works, but what we do with ourselves when it does.

One Thing Worth Noting

He built DeepMind in London, not Silicon Valley. Deliberately. Less competition for talent, more distance from the trend cycle, more space for 20-year thinking.

Being slightly outside the maelstrom, he says, is conducive to thinking deeply about things. You don't want to be distracted by the latest fad when your mission is generational.

That's advice that applies well beyond AI.