Alexander Lewis
Dec 20, 2023
In a groundbreaking leap for artificial intelligence, researchers at the Technical University of Denmark (DTU) have unveiled a pioneering AI system, trained on the life stories of over a million individuals, capable of remarkably accurate predictions about people's lives and their likelihood of early mortality.
The innovative AI, named "life2vec," was trained on extensive personal data from Denmark's population. By analysing health and employment records from 6 million Danes over 12 years, including education, medical history, income, and occupation, scientists transformed this wealth of information into a language-based model similar to the technology behind popular AI apps like ChatGPT.
Published in the journal Nature Computational Science, the study revealed that once the AI learned the intricate patterns within the data, it surpassed existing systems. The AI could predict outcomes such as personality traits and even estimate the time of death with striking accuracy.
In a particular test involving a group aged 35 to 65, half of whom passed away between 2016 and 2020, the AI's predictions outperformed other models by 11%. Lead researcher Sune Lehman emphasised the significance of viewing human life as a sequence of events, just like sentences in a language.
"While transformers in AI are typically employed for various tasks, in our experiments, we've used them to decipher what we term 'life sequences' — events within human lives," explained Dr Lehman.
The AI model yielded insights into survival probabilities within four years, revealing that factors like leadership roles and higher income correlate with increased survival chances. However, being male, skilled, or having mental health diagnoses was associated with a higher risk of mortality.
