Britain can lead the way in AI-powered medical advances: David Richards
Professor José Penadés had puzzled for years over how superbugs swap resistance genes. Last year he fed the question into Google’s ‘AI co-scientist’, a Google Gemini-powered system built to think alongside scientists. The machine took two days.
When Penadés saw the output he telephoned Google, convinced they had read his unpublished work. They had not.
The AI had reached the same conclusion his team had after years of work, and proposed a hypothesis his lab had not even considered. "I feel this will change science, definitely,” he told the BBC.
At the time this felt like a curiosity. Now it looks like a draft of how medicine will be done for a generation.
On April 28, the FDA announced that two cancer trials will report clinical endpoints and safety signals to the regulator in real time. AstraZeneca’s Phase 2 TRAVERSE trial in mantle cell lymphoma, running at MD Anderson and the University of Pennsylvania, is one. The other is Amgen’s Phase 1b STREAM-SCLC study in small-cell lung cancer, sites still being finalised.
For 60 years, as Commissioner Dr Marty Makary admitted last week, drug trials have stuck to a stop-start tempo, with whole years lost in the gaps between phases. In oncology that lag is no small thing, because cancer mutates while you study it, and a five-year trial often delivers its verdict on a target that has already moved on. Chief AI Officer Jeremy Walsh reckons the new approach could shave 20, 30 or even 40 per cent off trial time without compromising safety. Set that against the often-cited $2.6 billion cost of bringing a drug to market and you see why pharma will pay attention.
What is striking is that compression is happening at every stage of medical discovery at once.
Hypothesis generation that consumed a decade is collapsing to days. Trial evidence that once moved in batches is beginning to stream. And the regulator’s own scientific reviews, once a three-day exercise for a senior reviewer, are now done in minutes by the FDA’s internal AI.
None of those alone would shift the world. Together they amount to something nobody yet has a playbook for.
I have argued before that AI will eliminate 40 to 60 per cent of white-collar work inside three years, and stand by that. Dario Amodei of Anthropic, whose company builds these systems, has warned, similarly, that half of entry-level white-collar jobs could vanish in one to five years.
The trainee accountant whose first three years were ledger work, the paralegal still cutting her teeth on discovery, the audit associate, the compliance officer at the back of the open-plan office: each is in the line of fire. Westminster is still treating this as a problem for the next parliament rather than the current one.
But weigh that loss honestly against what is suddenly within reach: cancers caught while still small enough to remove cleanly; failed drugs killed before they devour another billion pounds; antibiotics designed in the lab to outpace the next superbug rather than chase it years late.
Britain happens to hold most of the cards needed: AstraZeneca, DeepMind, the Crick, the Sanger, and an NHS whose patient records, for all our complaints, run deeper and further back than almost anywhere on earth. We have every part needed. What we lack is the will to use them.
Get the next ten years right and our grandchildren may look back on the cancers and dementia that frighten us now the way we look back on smallpox and polio. That is the prize.
The technology poised to displace tens of millions of white-collar workers could also end forms of suffering humanity has dragged with it for as long as we have written history. We do not get to pick one half of that bargain. The question is not whether this is coming, but whether we steer it or merely watch.
David Richards MBE is managing partner of Yorkshire AI Labs