Life2vec, an artificial intelligence (AI) system, has deduced the lifespan of Danish citizens. This prediction is based on a comprehensive collection of data pertaining to individual’s lives, a process that offers a meticulous analysis of society.
An AI system named life2vec has collected a wealth of detailed data about individual’s life trajectory.
Using this rich data, it forecasts a person’s lifespan in a more accurate way than previous models, specifically among the Danish population.
The AI’s database includes information on Danes’ health, education, employment, income, residences, and more. This data was assembled by Statistics Denmark, according to a press release.
Life2vec was trained by Northeastern University in Boston, USA, the Technical University of Denmark, and the University of Copenhagen, as reported by Science Alert.
Life2vec was built using large AI language models.
One such language model is the popular AI, ChatGPT, but there are others that have also been made public.
In the future, Life2Vec’s information will only be shared with researchers. The developers assert that the model is merely used as a prediction tool.
The model has data on nearly all Danish citizens, amounting to about six million people.
“The research on this language model was conducted in Denmark, a country with its own unique culture, laws, and societal norms,” says researcher Tina Eliassi-Rad.
“This tool is like a watchtower for one society – not all societies.”
Eliassi-Rad is a computer science professor at Northeastern University.
“These tools allow you to perceive your society from a new perspective: what policy rules and regulations do we have? You can consider the model as a sounding board for what’s happening in Denmark.”
The model was unveiled in December by the science magazine Nature Computational Science.
In the study, the AI statistically monitored and predicted the lives of 2.3 million people aged 35-65, according to CNN.
This particular age group was selected as mortality in this range is typically harder to predict.
The model used Danish statistical data from 2008–2016. The researchers then tested the model by letting a counting program individually predict who survived past 2016.
The calculation program was correct in 78 percent of the cases.
Life2vec significantly outperformed other models in predicting mortality.
There may have been demographic distortions in the sample.
“For instance, if someone doesn’t have income data or doesn’t want to engage with healthcare, we don’t have access to their information,” the researchers mention.
It is noted by the Danes that such predictions have long been made in welfare states, particularly by insurance companies. The more accurate the model, the more assured the profitability of the insurance company.
Developers and Danes hope that Life2Vec will introduce a more human-centered approach to AI development.
“For instance, the tool should not be used to predict the lives of individual people,” Eliassi-Rad cautions.
Researchers view their model not as a finished product, but as a starting point for discussion. The question is, what can precise forecasts do, and how should they be utilized?
The model uses information it has gathered from the coincidences of millions of lives. A person’s life is depicted as a narrative. It’s akin to a lengthy sentence about the things that can occur in a person’s life.
Despite the model’s high degree of accuracy in making predictions, they are exclusively based on correlations. The model cannot predict the exact age of death or the cause of death for a person.
In cases where the probability of dying is low, the causes of death are those that the model cannot predict. These include, for instance, car accidents.
