Predicting Individual Life Outcomes Using Machine Learning, including Mortality
Danish researchers utilize advanced machine learning algorithms to make precise predictions about various aspects of individuals' lives, including their potential lifespan
A team of Danish researchers have harnessed the power of advanced machine-learning algorithms to successfully forecast various aspects of human lives, such as the likelihood of an individual's early mortality. The findings, featured in the recent issue of Nature Computational Science, expound on the capabilities of a machine-learning model known as life2vec, demonstrating its ability to anticipate the course of a person's life and behavior based on highly detailed personal data.
Speaking at a meeting held at Station F in Paris on May 26, 2023, OpenAI CEO Sam Altman reassured the audience that the company's technology, including the popular ChatGPT bot, would not lead to widespread job displacement. Altman sought to allay concerns about the impact of artificial intelligence (AI) on the job market during his address in Paris.
Joel Saget/AFP/Getty Images
OpenAIs technology is upending our everyday lives. Its overseen exclusively by wealthy, White men
"We have the ability to make predictions of any kind," stated Sune Lehmann, the principal investigator of the study and a professor at the Technical University of Denmark. It should be noted, however, that the current version of the model is a "research prototype" and is not capable of carrying out "real-world tasks."
Lehmann and his colleagues leveraged data from a national registry in Denmark that encompassed a varied population of 6 million individuals. The dataset included information spanning from 2008 to 2016 examining key facets of life, including education, health, income, and occupation.
The researchers used language processing techniques to create a vocabulary for life events, enabling life2vec to analyze sentences like "In September 2012, Francisco received twenty thousand Danish kroner as a guard at a castle in Elsinore" or "During her third year at secondary boarding school, Hermione followed five elective classes." According to Lehmann, the algorithm then learned from this data and could predict various aspects of individuals' lives, including their thoughts, emotions, behavior, and even the likelihood of death in the near future.
The team utilized data from January 1, 2008 to December 31, 2015 for a group of over 2.3 million individuals aged 35 to 65 in order to forecast potential early mortality. This age group was chosen due to the challenges in predicting mortality within that range, according to Lehmann. Life2vec then utilized the data to estimate the likelihood of an individual surviving the four years following 2016.
Lehmann explained, "In order to evaluate the effectiveness of [life2vec], we selected a sample of 100,000 individuals, with half surviving and half not. The researchers had access to the data on who had passed away after 2016, but the algorithm did not. The researchers then proceeded to test the algorithm's accuracy by having it predict whether each individual had survived past 2016. The results were remarkable, with the algorithm making correct predictions 78% of the time."
Life2vec also outperformed other state-of-the-art models and baselines by at least 11% by predicting mortality outcomes more accurately, the report said.
On March 8 in the United States, Chairman Mark Warner, D-Va., greets FBI Director Christopher Wray and ranking member Sen. Marco Rubio, R-Fla., as they welcome witnesses to the Senate Select Intelligence Committee hearing on worldwide threats in the Hart Building. The witnesses included National Security Agency Director Gen. Paul Nakasone, Director of National Intelligence Avril Haines, CIA Director William J.Burns, and Defense Intelligence Agency Director Lt. Gen. Scott D.Berrier. (Photo: Tom Williams/CQ Roll Call via AP Images)
US senators are suggesting imposing hefty penalties for AI-powered securities fraud and market manipulation. A study also discovered that being male, working as a skilled professional, or having a mental health diagnosis could lead to a higher risk of premature death, while holding a managerial position or having a higher income often correlated with increased chances of survival.
The study had some limitations. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment. Additionally, the researchers only analyzed data from an eight-year period, potentially introducing sociodemographic biases in the sampling despite the inclusion of every person in Denmark in the national registry.
"If an individual does not receive a salary, or opts out of participating in healthcare systems, their data is not accessible to us," they explained.
The research was carried out in a prosperous nation with a robust infrastructure and healthcare system, the authors also pointed out. It remains uncertain whether the discoveries of life2vecs can be extrapolated to other countries such as the United States, considering their economic and societal disparities.
Lehmann acknowledges that the algorithm may sound "ominous and crazy," but emphasizes that there has been significant research and development on it, particularly driven by insurance companies.
Dr. Arthur Caplan, the head of the Division of Medical Ethics at New York University's Grossman School of Medicine, agrees that insurance companies will be motivated to stay ahead of consumer demand as models like life2vec become more commercially available.
"This will complicate selling insurance in the future," he explained. "It's impossible to insure against risk when everyone is aware of the specific risks."
On the other hand, Caplan, who was not part of the new study, points out that life2vec does not forecast the age or cause of a person's death. For instance, an algorithm cannot predict if someone will die in a car accident."
Get CNN Health's weekly newsletter
Sign up here to get The Results Are In with Dr. Sanjay Gupta every Tuesday from the CNN Health team.
In five years, Caplan anticipates the emergence of more sophisticated prediction models. "We will have improved models with larger databases that will offer recommendations on how to extend your lifespan," he stated.
Caplan ultimately believes that using artificial intelligence to predict our death removes the element of mystery that makes life interesting. He expressed concern about the potential for robots to manipulate information and predict so much about our behavior that our lives become too predictable, diminishing their value. He emphasized that instead of worrying about robots taking over the world, we should be concerned about the impact of their ability to foresee our actions.
