“I see your death” – gloomy prediction algorithm

Researchers in Denmark say they have used powerful machine learning algorithms to predict a person's behavior and even death.

Their study, published this week in the journal Nature Computational Science, shows how a machine learning algorithm model, called life2vec, predicted a person's life outcomes and actions when given very specific data about the same thing.

Prediction of any kind

Using this data, “we can make all kinds of predictions,” said Sonny Lehmann, lead author of the study and a professor at the Technical University of Denmark. However, the researchers note that it is a “research prototype” and cannot perform “real-world tasks” in its current state.

Lehmann and the other authors of this study used data from a national registry in Denmark detailing a diverse group of 6 million people. Specifically, they collected information from 2008 to 2016 related to important aspects of their lives, such as education, health, income, and occupation, the report writes. CNN.

What is life2vec

The researchers adapted language processing techniques and created a vocabulary for life events so that life2vec could interpret sentences based on data, such as “In September 2012, Francisco received twenty thousand Danish kroner as a guard at a castle in Elsinore” or “While in her third year of high school,” Hermione took five electives.

The algorithm then learned from that data, Lehman explains, and was able to make predictions about certain aspects of people's lives, including how they might think, feel and behave, and even if they would die within the next few years.

Predicting death

To predict how early someone might die, the team used data from January 1, 2008 to December 31, 2015 for a group of more than 2.3 million people between the ages of 35 and 65. Lehman said this group was chosen because deaths in this age group are difficult to predict.

Life2vec used the data to infer a person's likelihood of surviving the next four years after 2016. “To test how good (life2vec) is, we selected a group of 100,000 people where half survived and the other half died,” Lehman said. The researchers identified people who died after 2016, but the algorithm did not. He was then placed under surveillance. They had the algorithm make individual predictions about whether or not someone lived after 2016. The results were impressive: the algorithm was correct 78% of the time.

Managerial positions and longevity

The report concluded that Life2vec also outperformed other recent models and baselines by at least 11 percent, more accurately predicting mortality outcomes. Notably, men were more likely to die after 2016. The researchers found that occupations such as working as a skilled worker, such as an engineer, or having a diagnosed mental health problem, such as depression or anxiety, also lead to early death. Meanwhile, managerial positions and higher incomes ensure that people live longer, the same study says. However, the research had several limitations, both from the time it was conducted, 8 years old, but also because it was conducted in a rich country with strong infrastructure and a strong healthcare system, among other individual criteria, adds Lehmann.

Finally, the statement of Dr. Arthur Caplan, chair of the department of medical ethics at New York University's Grossman School of Medicine, notes that insurers will race to outpace consumers as models like life2vec become more commercial. “This will make it harder to sell insurance in the future, because you simply can't do risk insurance if everyone knows exactly what the risks are,” he concluded, predicting that within five years more advanced predictive models will come together. Using artificial intelligence, they will remove from human life the element that makes it most interesting: mystery.