WATCH: The future of British healthcare: Nightingale AI
Aldo Faisal, Professor of AI & Neuroscience and Co-Director of the School of Convergence Science at Imperial College London, has long been inspired by Star Trek: Voyager. Its holographic medical officer, known simply as The Doctor, is programmed with the knowledge of 47 medical officers, 3,000 alien cultures and over 5 million distinct medical treatments.
“That’s always stuck with me,” he says. “Why can’t we have that in reality rather than just science fiction? Well, maybe we can.”
His team’s answer is Nightingale AI, which they believe is the world’s largest health foundation AI model. They named it after Florence Nightingale, the 19th-century pioneer of modern nursing. It is a clinical intelligence system that can support an individual’s care from initial triage through to highly personalised treatment recommendations, always alongside a human clinical team.
The problem
Health is rarely simple. Ask an oncology consultant and they’ll tell you that no two cancer patients are alike. With differing medical histories, predicting how each will respond to a particular drug is complex and time-consuming.
Neurologists face a different problem. When the brain deteriorates it does so rapidly, and the warning signs can be invisible even to highly trained clinicians. Earlier insight could prevent irreversible secondary brain injuries.
In a busy intensive care unit, clinicians often hold all the pieces of the puzzle for a patient but must assemble them by hand, at the most stressful moments.
More than words
Large language models such as ChatGPT and Claude reason using written language alone, whereas human health involves far more data than text.
Nightingale AI is multi-modal model, designed to learn from electronic medical records, lab results, X-rays, genetic profiles and published medical research. From this, it aims to build specific insight into human health and, ultimately, predict how an individual patient’s medical journey is likely to unfold. That gives the clinical team the power to choose the best interventions.
Built on the NHS
The project is possible because the NHS is three to four years ahead of the world in digitising medical records. As a high-priority UK government project, Nightingale AI has launched with 60,000 public-domain records. The plan is to scale to all 65 million people in the UK. Aldo expects the medical records of all 450 million people on the continent to be digitised by 2029, and the team’s goal is to draw on the records of half a billion patients by 2030.
At that scale, the applications could change everything. Nightingale AI could:
- Power an automated triage service, such as NHS 111, for instant, accurate assessment
- Create a personalised care pathway for every patient, based on their own medical history and demographic
- Act as an AI bedside clinician, assisting doctors in real time in oncology, neurology and intensive care, for adults and children
- Speed up drug development
As we know by now, AI is very good at spotting patterns. By monitoring huge datasets of real medical records, it could reveal beneficial effects of existing drugs that we don’t yet know about. Examples of this already exist, such as Viagra, which was originally developed for heart disease and angina.
Isambard-AI powering Nightingale AI
Processing data at this scale takes masses of compute time and Nightingale AI has so far been awarded one million GPU hours on Isambard-AI, the UK’s fastest AI supercomputer, hosted at the University of Bristol.
“Isambard-AI is absolutely critical to this project,” says Aldo. “Because it is a sovereign UK resource, we can analyse sensitive NHS data without it ever leaving UK shores.”
The results speak for themselves. By using Isambard-AI to refine their multi-modal coding techniques, the team has already achieved a 40x increase in the efficiency of their foundation model code.
“Without Isambard-AI, Nightingale AI wouldn’t be possible,” Aldo says. “It has completely supercharged our development, and we expect our ongoing access to develop the model even further.”
“Together with Isambard-AI,” Aldo says, “we are making the idea of a truly personalised healthcare, once just the realm of science fiction, into a reality for us all.”