A groundbreaking new artificial intelligence system, similar to ChatGPT, trained on the life stories of over a million people in Denmark, has demonstrated high accuracy in predicting individual life outcomes and the risk of early death, according to a study by scientists from the Technical University of Denmark (DTU).
The AI model, named “life2vec,” was trained on personal data, including health and labour market information, for 6 million Danes from 2008 to 2020. The dataset was transformed into words to train the language model, and it outperformed existing systems in predicting outcomes such as personality and time of death, as per a report by the Independent.
The study, published in the journal Nature Computational Science, found that the model’s predictions were 11 per cent more accurate than other AI models and methods used by life insurance companies.
The researchers applied the model to a group aged 35 to 65, half of whom died between 2016 and 2020, and found that it surpassed other models in accuracy. The study’s lead author, Sune Lehman from DTU, highlighted the model’s ability to analyze life events and sequences, likening human life to a long sequence of events, similar to a sentence in language.
Once the AI model grasped the patterns in the data, it kicked things up a notch. Outshining other advanced systems, it showcased an impressive ability to predict outcomes like personality traits and even the time of death with remarkable accuracy.
It revealed associations, such as individuals in leadership positions or with high income being more likely to survive, and factors like being male, skilled or having a mental diagnosis being linked to a higher risk of death.
While the model showed promise in predicting outcomes and personality traits, scientists emphasized ethical concerns and cautioned against its use by life insurance companies.
Dr Lehman explained that the model’s predictions could conflict with the fundamental idea of insurance, which relies on sharing the lack of knowledge about who may be affected by unexpected incidents.
The researchers believe that the model’s capability to provide precise answers opens avenues for understanding new potential mechanisms impacting life outcomes and exploring personalized interventions.
(With inputs from agencies)