Scientists have developed an algorithm that uses a person's life story to predict how he will live and when he will die.
According to a new study, the model called 'life2vec', iIt is accurate about 78 percent of the time, which puts it on par with other algorithms designed to predict similar life outcomes.
But unlike other models, it works like a chatbot, using existing details to predict what comes next.
It was built by scientists in Denmark and the US who trained a machine learning algorithm on a huge amount of Danish data, giving it all kinds of information about more than six million real people, including income, occupation, place of residence, injuries and pregnancy . history.
Death knell? The new tool called life2vec can predict the likelihood that someone will die within a certain period of time, but don't worry: the data is not available to the public
Their end result was a model that can process ordinary language and generate predictions about a person's likelihood of dying early, or about their lifetime income.
Some of the factors that can lead to an earlier death include being male, having a mental health diagnosis, or working in a skilled trade. Things associated with a longer life include higher income or a leadership role.
By viewing each part of your life as if they were words in a sentence, life2vec predicts where the story will go based on what has been written so far.
Just as ChatGPT users ask it to write a song, poem or essay, scientists can ask life2vec simple questions like “dead in four years?” for a specific person.
The model was trained on data from 2008 to 2016.
Based on their population data, the country correctly predicted who had died in 2020 in more than three-quarters of the cases.
The research appeared in Natural Computational Sciences.
However, to protect the personal information of the people whose data was used to train the system, it is not available for the general public (or companies) to use, lead researcher Sune Lehmann told DailyMail.com.
Based on data about mi
“We are actively working on ways to share some results more openly, but this will require further research in a way that can guarantee the privacy of the people in the study,” said Lehmann, professor of networks and complex systems at the Technical University of Denmark.
Even if the model is finally available to the public, Danish privacy law would make it illegal to use life2vec to make decisions about individuals, such as writing insurance policies or making hiring decisions.
In much the same way that ChatGPT and other large language models are trained from existing written works, life2vec was taught by data from people's lives, written out as a series of data-rich sentences.
These include sentences such as “In September 2012, Francisco was paid twenty thousand Danish crowns as a guard at a castle in Elsinore” or “During her third year at secondary boarding school, Hermione took five electives.”
Lehmann and his team assigned different tokens to each piece of information, and these pieces of data were all mapped in relation to each other.
Categories in people's life stories span the range of human experiences: a forearm fracture is shown as S52; working in a tobacconist is coded as IND4726, income is represented by 100 different digital tokens; and 'pospartum haemorrhage' is O72.
Many of these relationships are intuitive, such as occupation and income: certain jobs pay more money.
But what life2vec does is map out the vast constellation of factors that make up an individual's life, allowing someone to ask them to make a prediction based on millions of other people and many factors.
It can also make predictions about people's personalities.
To do this, Lehmann and his team trained the model to predict people's answers to questions on a personality test.
The test asks respondents to rate 10 items based on their level of agreement, items such as “The first thing I always do in a new place is make friends,” or “I rarely give my opinion during group meetings.”
It's important to note, Lehmann said, that the data all came from Denmark, so these predictions may not apply to people living in other places — in addition to the fact that most people probably don't really want to know when they will die.
“The model opens up important positive and negative perspectives to discuss and tackle politics,” says Lehmann told Newswise.
'Similar technologies for predicting life events and human behavior are already being used today within technology 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.
“This discussion must be part of the democratic conversation so that we can think about where technology is taking us and whether this is a development we want.”