Dario Amodei says AI will cure cancer and Alzheimer's in a decade. Doctors aren't buying it.
By Saghir Hussain ·
Dario Amodei says AI will cure cancer and Alzheimer's in a decade. I want to believe him. I really do. But I've heard this tune before — usually from men in Silicon Valley hoodies who've never sat in an A&E waiting room in Blackburn or watched a loved one fade on a Glasgow dementia ward.
The promise is seductive. The timeline is fantasy.
Anthropic's chief executive told the Business Insider this week that compressed 21st-century biology — AI designing drugs, simulating cells, cracking protein folding — will deliver cures for cancer and Alzheimer's within ten years. He calls it "compressed 21st century." Doctors call it rubbish. The Royal College of Physicians didn't put it quite that bluntly, but their response amounted to the same thing: show us the clinical trials, not the pitch deck.
Amodei isn't a charlatan. Anthropic's Claude models are genuine breakthroughs. But there's a category error here. AI can accelerate discovery. It cannot replace the messy, regulated, human reality of medicine. You don't cure Alzheimer's in a simulator. You cure it in Phase III trials that take years, cost billions, and fail 90% of the time. No amount of compute changes that.
This isn't abstract for British patients
While Amodei talks timelines, the NHS is buckling. Cancer waiting lists in England hit 360,000 last month. The 62-day treatment target hasn't been met since 2015. In Wales, some patients wait over a year for a diagnosis. Alzheimer's? There are 900,000 people living with dementia in the UK. Most don't need a theoretical cure in 2034. They need a care package this afternoon that isn't being rationed by a cash-strapped council.
AI is the most important technology of our lifetime. I've said it repeatedly in these pages. But we are moving faster than our safeguards. The MHRA is still figuring out how to regulate AI-as-a-medical-device. The NHS AI Lab has pilots, not pathways. And the data? Our patient records are fragmented across trusts that don't talk to each other, guarded by procurement rules written for fax machines.
Meanwhile, the money flows one way. Anthropic's valuation north of $18 billion. UK life sciences investment? A fraction of that, and half of it clustered in the Golden Triangle. Try raising a Series A for a diagnostics startup in Middlesbrough or Swansea.
The safeguards aren't keeping up
We need a regulatory framework that matches the speed of the science. Not red tape — smart tape. Conditional licensing for AI-discovered compounds with real-world evidence collection built in. A national health data architecture that lets researchers train models on the whole population, not just the 15% who consent to research. And a legal liability regime so that when an AI-designed drug harms someone — and one will — we know who answers for it.
Amodei's optimism isn't the problem. Unmoored optimism is. The cures will come. AI will transform medicine. But the next decade belongs to the boring work: integration, validation, regulation, equity. The work that doesn't make headlines but determines whether a grandmother in Bradford gets the same shot at a breakthrough as a venture capitalist in Palo Alto.
So by all means, cheer the science. Just don't mistake the press release for the prescription.
When the first AI-designed Alzheimer's drug clears Phase III, I'll buy the champagne. Until then, I'll take a funded social care system and a functioning NHS.