Danish Researchers Build AI Model Aimed To Predict Life

Researchers in Denmark are using artificial intelligence (AI) and a lot of data from many people to predict different stages of someone’s life, including the end. Their project, called life2vec, aims to show how powerful this technology can be but also to warn about its risks. The researchers want to use advanced computer programmes to find patterns that can forecast various health or social events in a person’s life.

Sune Lehmann, a professor at the Technical University of Denmark and co-author of a study, explained to AFP that their framework is designed to make a wide range of predictions about people’s lives, as long as there is data to learn from. It can forecast health-related events like fertility, obesity, or cancer risk, as well as other aspects such as financial success. The system works similarly to ChatGPT but focuses on analysing life factors like birth, education, social benefits and work patterns.

The research team is applying the technology behind language-processing algorithms, like the ones used in ChatGPT, to study and predict human life events. They see a person’s life as a series of events, such as birth, school, moving and marriage. However, the announcement of their program led to some people wrongly calling it a “death calculator.” This misunderstanding caused fake websites to trick individuals into sharing personal information in exchange for a life expectancy prediction. The researchers have stated that their software is currently kept private and not accessible to the public or other researchers.

The life2vec model uses anonymous data from about six million Danish people, provided by Statistics Denmark. This model can predict different life outcomes, including when someone might die, by looking at the sequence of events in their lives. It has a 78% accuracy rate for predicting death and a 73% accuracy rate for predicting if someone will move to a new city or country.

The researchers focus on predicting early mortality in people aged 35 to 65. They use data from eight years (2008 to 2016) to forecast whether someone will pass away in the following four years. According to Lehmann, this model predicts death more accurately than any other algorithm they have compared it with.