A new AI model can now take lots of information about you, and predict with 78% accuracy when you will likely die, according to a new study conducted by researchers in Denmark and the U.S.
Lead author Sune Lehmann said the model, called Life2vec, uses the same ‘transformer‘ technology behind ChatGPT “to analyze human lives by representing each person as a sequence of events that happens in their life.”
It then uses that information to predict the time of death of people in the study. The research examined aspects of a person’s life story between 2008 and 2016 and was published this week in Nature Computational Science.
Also read: 2023: The Year Generative AI Shined Bright—and Cast a Shadow
AI says males die early
Researchers say the AI model, dubbed “the doom calculator” by the Daily Mail, was trained on the data of six million people from Denmark. It looked at things like their age, health, education, jobs, income, injuries and other life events.
Life2vec was taught to understand information about the lives of people in sentences like “In September 2012, Francisco received 20,000 Danish kroner as a guard at a castle in Elsinore.” Or, “During her third year at secondary boarding school, Hermione followed five elective classes,” researchers said in the study.
Eventually, the AI model was able to construct “individual human life trajectories” and to accurately predict those who had died by 2020 about 78% of the time.
The “model can predict almost anything,” including personalities and individual earnings, according to Lehmann, a professor of networks and complexity science at the Technical University of Denmark. He worked with computer science professor Tina Eliassi-Rad from Notheastern University in Boston, U.S.
“The whole story of a human life, in a way, can also be thought of as a giant long sentence of the many things that can happen to a person,” Lehmann said on the university news site.
None of the people who took part in the study were informed of their death predictions. Lehmann told the New York Post that doing so “would be very irresponsible.”
So far, Life2vec has been tested on people aged between 35 and 65 in Denmark, half of whom are now dead. People that earn more money and those in leadership roles were found to live longer. But being male, a smoker, having a mental health diagnosis or having a skilled profession was linked with dying early.
Our wild new paper "Using sequences of life-events to predict human lives" is finally out!!
Check it out here: https://t.co/0OyKvUNuA1 pic.twitter.com/fxAcmDornj
— Sune Lehmann (@suneman) December 18, 2023
Doom calculator closed to the public – for now
Sune Lehmann said that Life2vec will not be made available to the general public or companies for now in order to protect the privacy of those people whose information was used to train the AI.
“We are actively working on ways to share some of the results more openly, but this requires further research to be done in a way that can guarantee the privacy of the people in the study,’ he told the Daily Mail.
And even when the model is finally available to the public, it might still not be usable outside of Denmark. That’s because local privacy laws would prevent the AI from being “used to make decisions about people – like writing insurance policies.”
“This kind of tool is like an observatory of society – and not all societies,” said co-author Eliassi-Rad. “Whether this can be done in America is a different story.”
She added that tools like Life2vec should not be used to predict the outcomes of individual people [because most people probably do not really want to know when they will die], but instead to track the different trends in society.
“Even though we’re using prediction to evaluate how good these models are, the tool shouldn’t be used for prediction on real people,” she told Notheastern Global News, the university news site. Real people “have hearts and minds.”
Lehmann said “the model opens up important positive and negative perspectives to discuss and address politically,” as reported by Newswise.
“Similar technologies for predicting life events and human behavior are already used today inside tech companies that, for example, track our behavior on social networks, profile us extremely accurately, and use these profiles to predict our behavior and influence us.”
Complicated road ahead
Some experts are worried about artificial intelligence getting involved in estimating death. Art Caplan, professor of bioethics at New York University Langone Medical Center in New York City, said this is “the start of a very complicated road.”
“These algorithms are starting to take away things we normally don’t know,” Caplan told USA Today. “It has upside and could prevent deaths, but it’s got a real existential threat of taking all the unknowns out of life, which is not necessarily a good thing.”
There are other scientists are building systems around “blood and other physical and medical features to make predictive forecasts,” similar to what happens in the insurance business.
and then