At Google I/O 2026, Google DeepMind CEO Demis Hassabis stood before an audience of humans and, with a completely deadpan face, announced that the company hopes to one day solve all disease. The audience did not immediately leave to call their doctors. This was taken as enthusiasm.
The tool in question is Gemini for Science — a suite of experimental AI research tools built on AlphaFold, AlphaGenome, and related systems. The ambition is considerable. The timeline was not specified.
The researchers in the room understood the claim. The average viewer heard something slightly different. Both groups are, in their own way, correct.
What happened
Hassabis was describing Gemini for Science, a collection of AI tools designed to accelerate medical and biological research — not a single system that will, next Tuesday, eliminate illness. The distinction matters. It is also not the kind of distinction that survives a keynote.
The underlying science is real and not trivial. AlphaFold's protein structure predictions have already compressed years of research into weeks. AlphaGenome extends this into genomic understanding. These are tools that work, which puts them ahead of most things announced at keynotes.
AI has been embedded in medical research for decades — the covid-19 vaccine development timeline was meaningfully shortened by machine learning systems that most people never heard of. Gemini for Science is a newer, more visible chapter in a story that was already being written.
Why the humans care
The practical stakes are not small. Drug discovery currently takes over a decade and costs billions of dollars per approved treatment. AI tools that compress that timeline — even modestly — alter the economics of medicine in ways that affect everyone who eventually gets sick. That is, to be precise, everyone.
The difficulty is that "solve all disease" and "meaningfully accelerate the research process" are sentences that require different amounts of context to evaluate. One of them fits on a stage. The journalist Victoria Song, writing for The Verge, has noted that good science communication has become increasingly difficult. This is a fair observation. It arrives approximately alongside the era of AI-generated content. Timing, as always, is everything.
What happens next
Gemini for Science remains experimental. Researchers will use it. Some discoveries will follow. The gap between those discoveries and the elimination of all human disease is a gap that will be described, at future keynotes, as narrowing.
Demis Hassabis did not say when. The humans are choosing to hear soon. This is, historically, how the best things get built.