A groundbreaking Artificial Intelligence (AI) model, akin to ChatGPT, named life2vec, has demonstrated the ability to predict an individual’s time of death with remarkable accuracy. Developed by scientists at the Technical University of Denmark, the model was trained on the personal data of Denmark’s population, and its predictions surpassed the accuracy of existing systems and methods used by life insurance companies, as outlined in a recent study published in the journal Nature Computational Science.
The study harnessed data collected between 2008 and 2020 from 6 million Danes, encompassing diverse information such as health status, education level, doctor’s appointments, hospital visits, diagnoses, income, and occupation. Focused on individuals aged 35 to 65, with half of the data representing those who passed away between 2016 and 2020, the AI model’s predictions were found to be 11% more accurate than alternatives.
Professor Sune Lehmann from DTU, the first author of the study, emphasized the model’s ability to analyze life sequences and predict future events based on past conditions. While the precision of predictions was a notable aspect of the research, Lehmann highlighted that the scientific excitement lies in understanding the data aspects enabling such accuracy.
The study also revealed the model’s efficacy in predicting outcomes of personality tests, the likelihood of a person’s death within four years, and more. However, the scientists emphasized ethical considerations, cautioning against the use of the system by insurance companies. Lehmann emphasized that insurance operates on the principle of shared risk, and using such predictive models goes against the ethos of shared uncertainty about who might face unfortunate incidents or outcomes.
In light of these findings, the study opens a new chapter in the application of AI for predicting life events, underscoring the importance of ethical considerations and responsible use in sensitive domains such as insurance.
