In an effort to increase public awareness of the potential benefits and risks of technology, researchers in Denmark are using artificial intelligence and data from millions of people to predict a person’s life events from the beginning to the end.
The developers of life2vec, far from being drawn to macabre fascinations, aim to investigate patterns and linkages that so-called deep-learning programs might unearth to forecast a broad spectrum of social or health-related “life-events”.
It’s a fairly broad framework for forecasting the lives of people. One of the authors of a study that was just published in the journal Nature Computational Science, Sune Lehmann, a professor at the Technical University of Denmark (DTU), told AFP that “it can predict anything where you have training data.”
For Lehmann, the options are virtually limitless.
It might forecast medical results. Thus, it might be able to forecast obesity or fertility, as well as the likelihood of developing cancer or not. However, it can also indicate whether you’ll be wealthy,” he added.
The program employs a methodology akin to ChatGPT’s, but instead examines life-influencing factors like birth, education, social advantages, and even work schedules.
The group is attempting to “examine the evolution and predictability of human lives based on detailed event sequences” by modifying the advancements that made language-processing algorithms possible.
“From one perspective, lives are simply sequences of events: People are born, visit the paediatrician, start school, move to a new location, get married, and so on,” Lehmann said.
However, the revelation of the program led to the emergence of a so-called “death calculator,” where deceitful websites tricked individuals into providing personal information in exchange for a prediction of their life expectancy using the AI program.
The researchers emphasize that the software is currently kept private and not accessible to the internet or the wider research community.
Six Million Data Points
The life2vec model is built upon the anonymised data of approximately six million Danes, which was collected by the official Statistics Denmark agency.
Through the analysis of sequences of events, one can make predictions about life outcomes, even up until the very end.
When it comes to predicting mortality, the algorithm has an accuracy rate of 78 percent. Similarly, when it comes to forecasting a person’s relocation, the algorithm has a success rate of 73 percent.
“We examine the phenomenon of early mortality.” We select a group of individuals in the age range of 35 to 65 years old. “We attempt to make predictions by analyzing data from an eight-year timeframe spanning from 2008 to 2016, to determine the likelihood of an individual’s mortality within the following four years,” explained Lehmann.
“The model excels in that area, surpassing any other algorithm we could find,” he said.
Researchers have chosen to focus on this specific age bracket, where deaths are typically infrequent, in order to validate the algorithm’s reliability.
Unfortunately, the tool is still in development and is not currently suitable for use outside of a research environment.
“Currently, it is a research project where we are investigating the realm of possibilities,” Lehmann stated.
He and his colleagues are also interested in investigating long-term outcomes, as well as the influence of social connections on life and health.
‘Public Counterpoint’
The project offers researchers a scientific alternative to the substantial investments made by large technology companies in AI algorithms.
“They can also build models like this, but they’re not making them public. They’re not talking about them,” Lehmann said.
“They are simply constructing them with the intention of increasing advertisement sales, either by selling more advertisements or by selling more products to you.”
He said it was “important to have an open and public counterpoint to begin to understand what can even happen with data like this”.
Pernille Tranberg, a Danish data ethics expert, highlighted the significance of this observation, emphasizing the existing utilization of comparable algorithms by various industries, including insurance companies.
“They probably put you into groups and say: ‘Okay, you have a chronic disease, the risk is this and this’,” Tranberg said.
“It can be used to unfairly discriminate against individuals, resulting in higher insurance premiums, denied loans, or limited access to public healthcare,” she said.
Developers have attempted to commercialize algorithms that predict our own demise.
“On the web, we’re already witnessing the emergence of prediction clocks that provide insights into our potential lifespan,” Tranberg remarked. “Certain individuals lack reliability.”
