European researchers have developed a groundbreaking artificial intelligence system capable of foreseeing the onset of over a thousand diseases many years before their actual manifestation. This innovative AI model bases its predictions on an individual`s personal medical history, as detailed in a pioneering study published in the esteemed scientific journal Nature.
The AI model, named Delphi-2M, is a sophisticated modification of the GPT (Generative Pre-trained Transformer) architecture, specifically adapted for simulating the progression of human diseases. It demonstrates remarkable accuracy in predicting the likelihood of a vast array of illnesses for each person, achieving results comparable to existing models that are specialized in predicting a single disease. Furthermore, Delphi-2M`s generative capabilities extend to creating hypothetical future health scenarios, offering an informative assessment of potential disease burden for up to two decades.
To train this powerful predictive tool, scientists utilized extensive datasets. They incorporated health information from 400,000 participants in the UK Biobank project, a comprehensive repository of biological samples and health records from approximately half a million individuals. Additionally, data from 1.9 million Danish citizens were used to further enhance the model`s training.
The researchers concluded that GPT models, due to their proven effectiveness in both predictive and generative tasks, are exceptionally well-suited for applications within the healthcare sector. Their potential is particularly significant for analyzing and leveraging large-scale national health databases, offering new avenues for preventative medicine and personalized healthcare.
This collaborative research effort involved scientists and institutions from Germany, the United Kingdom, Denmark, and Switzerland, highlighting a broad international commitment to advancing medical science through artificial intelligence.
