This AI can predict a person’s death with 78% accuracy

Do you want to know when you will die? You don’t need to ask this question to a doctor or a tarot reader: Artificial Intelligence has the answer. Danish researchers say they have used powerful machine learning algorithms to accurately predict certain aspects of human life, such as a person’s chance of dying early.

Their study, published in the journal Nature Computational Science, describes how a machine learning algorithm model called Life2Vac can predict the outcome of a person’s life and actions when presented with very specific data about them.

With that data, “We can make any kind of prediction”Sune Lehmann, lead author of the study and professor at the Technical University of Denmark, says in the conclusion of the work. But you don’t need to ring bells. The researchers say it is a “research prototype” and in its current state it cannot perform any “real-world functions.”

How does this work

Lehmann and his co-authors used data from a national registry in Denmark describing a diverse group of 6 million people. These included information from 2008 to 2016 related to important aspects of life such as education, health, income and occupation.

The researchers adapted language processing techniques and created a vocabulary of life events so that life2vec could interpret sentences based on the data, such as “In September 2012, Francis was given twenty thousand Danish crowns as a guard at a castle in Elsinore. Received” or “During” her third year at boarding school, Hermione took five elective classes.”

The algorithm then learns from that data, Lehman says, and is able to make predictions about certain aspects of people’s lives, such as how they think, feel and behave, and even Could the person die in the next few years?

Continue reading story

To estimate how quickly someone might die, the team used data from a group of more than 2.3 million people aged 35 to 65 from January 1, 2008, to December 31, 2015. This group was selected because mortality in that age group is more difficult to predict, explains Lehmann in the special journal article.

Life2Wake uses the data to estimate the probability that a person will live four years after 2016. “To test how well it (Life2Wake) works, we choose a group of 100,000 individuals, half of whom survive and half of whom die,” explains Lehmann. The researchers knew which people had died after 2016, but the algorithms did not.

After this, he tested it. They had algorithms that made personalized predictions about whether a person was alive before 2016. The results were impressive: the algorithm was correct 78% of the time, According to the report, Life2vec also outperformed other state-of-the-art models and baselines by at least 11% in predicting mortality outcomes more accurately.

Men were more likely to die after 2016. Being a skilled worker, such as an engineer, or being diagnosed with a mental health problem such as depression or anxiety also led to earlier death, according to the researchers. Meanwhile, being in a managerial position or having a high income often pushes people toward the “surviving” column.

However, the research had several limitations. “We observed that the experiments were not randomized and that the investigators were not blinded to allocation during the experiments and evaluation of the results,” the report said.

The researchers only analyzed data from an eight-year period, and there may have been socio-demographic biases in the sample, even though everyone in Denmark is listed in the national registry. “If someone doesn’t have a salary – or chooses not to participate in the health system –, we don’t have access to your data,” he explains.

Accurately predict death in 78% of cases

In the end, the AI ​​program was able to correctly predict who died in 2020 in 78% of cases, the researchers say in the report. None of the study participants were informed about their predicted death.

“That would be very irresponsible,” Lehman told the New York Post. Some of the factors associated with earlier death included having a mental health diagnosis, being male or having a skilled profession, The Science Times reported. Leadership roles in the workplace and higher income were associated with longer life expectancy. Lehman told the Post that the program can predict personalities and decisions to make international moves. “This model can predict almost anything,” he said.

When will I be able to enter my data into the death calculator?

Not so fast. The program and its data have not been made public to protect the privacy of those whose information was used. Lehmann told the Daily Mail, “We are actively working to share some of the results more openly, but this requires further research in a way that guarantees the confidentiality of study participants.” Give.”

Experts assure that its predicted power cannot extend beyond Denmark. “This type of instrument is like an observatory of society, not all societies,” he said. “Whether it can be done in the United States is a different story.” In any case, tools like life2vec should be used to monitor social trends, not to predict individual outcomes.

“Although we use predictions to evaluate the quality of these models, the tool should not be used to predict the evolution of real people,” he told the university’s news website. Real people “have hearts and minds.”

Lehman hopes the project will shed light on the evolution of AI and the need to make predictions, he told the university’s news website. “I don’t have those answers, but it’s time for us to start the conversation, because what we do know is that the detailed prediction of human life is already happening,” he explains via the same medium. “And right now there are no negotiations and it’s happening behind closed doors.”

More news that may interest you:

On video | They create a doll with artificial intelligence to tackle loneliness in older people