A team of scientists has developed an algorithm that uses an individual's life story to predict how long they will live and when they will die. The study revealed that the accuracy of the model named "life2vec" is approximately 78%, putting it on par with other algorithms designed to predict similar life outcomes. Researchers in Denmark and the United States trained the machine learning algorithm on a vast dataset from Denmark, which includes information about over six million people, including income, occupation, and place of residence.
They developed a novel model capable of processing simple language and generating predictions about an individual's likelihood of dying early or their income over their lifetime. Scientists can ask simple questions to life2vec, similar to how ChatGPT is used to write songs, poems, or articles. The model was trained on data from 2008 to 2016 and correctly predicted who would die by 2020 in more than three-quarters of the cases.
Lead researcher Sony Lehmann, a professor of networks and complex systems, stated that all the data came from Denmark, so the predictions may not apply to people living in other locations, alongside the fact that most people probably do not want to know when they will die. He added, "The model opens up important positive and negative perspectives for discussion and political handling." It should be noted that when the model is presented to the public, Danish privacy laws would make it illegal to use life2vec to make decisions about individuals, such as writing insurance documents or making hiring decisions. The study was published in the journal Nature Computational Science.
