PARIS:
In what could mark a turning point for modern medicine, European scientists have developed an artificial intelligence (AI) model capable of predicting future illnesses years before symptoms emerge.
The system, called Delphi-2M, was unveiled in a paper published in Nature on Wednesday. Drawing on the same “transformer” architecture behind consumer chatbots such as ChatGPT, the tool analyses medical histories to forecast the likelihood of more than 1,000 diseases.
“We’ve essentially taught the AI to read the grammar of healthcare,” explained Moritz Gerstung of the German Cancer Research Center. “It learns how diagnoses occur in sequence and combination, enabling very meaningful predictions about what might come next.”
To train the model, researchers tapped into Britain’s UK Biobank, a vast biomedical database holding details of half a million people. They then tested its accuracy using Denmark’s national health records covering nearly two million individuals.
Charts presented by Gerstung showed the AI flagging patients with unusually high or low risk of heart attacks compared with standard age-based predictions. Co-author Tom Fitzgerald of the European Molecular Biology Laboratory noted that such systems could one day “optimise resources across already stretched healthcare systems”.
Still, experts caution against premature hype. The datasets used — primarily British and Danish — were limited in terms of age and ethnicity, raising concerns about bias. “It’s a long way from improved healthcare,” warned Peter Bannister of the Institution of Engineering and Technology.
Even so, the tool’s potential is hard to ignore. Unlike current risk calculators such as Britain’s QRISK3, which focus narrowly on cardiovascular health, Delphi-2M can scan for thousands of diseases simultaneously and years in advance.
Ewan Birney, another co-author, said such breadth opens the door to “preventative medicine on a scale never seen before”. Gustavo Sudre, a professor of medical AI at King’s College London, added that the project was a “significant step towards scalable, interpretable and ethically responsible predictive modelling.”
The bigger picture
While not ready for clinical use, Delphi-2M hints at a future where healthcare shifts from reactive treatment to proactive prevention. Doctors could monitor high-risk patients earlier, intervene before crises, and relieve some of the pressure on overburdened systems.
