A new artificial intelligence system, similar to ChatGPT but developed using personal stories from over a million individuals, has shown remarkable accuracy in predicting individual life paths and the likelihood of premature death, as revealed by a recent study.
This artificial intelligence was programmed using personal records of all residents of Denmark. Researchers from the Technical University of Denmark (DTU) claim that it outperforms any current system in predicting death risks.
The team examined health data and labor market information for 6 million Danish citizens from 2008 to 2020, including details such as educational background, doctor visits, diagnoses, income, and job type.
They translated this information into texts to train an advanced linguistic model called “Life2vec”, using technology similar to that used in artificial intelligence applications like ChatGPT.
After identifying data patterns, the artificial intelligence was able to outperform other high-tech systems in predicting aspects such as personality traits and time of death with remarkable accuracy, as reported in the journal of Computational Natural Sciences this Tuesday.
The artificial intelligence was tested on a sample of individuals aged between 35 and 65, half of whom died between 2016 and 2020. The AI successfully predicted who would live or die with an 11% higher accuracy rate than any other artificial intelligence or methods used by life insurance companies.
Liemann from the Technical University of Denmark, the study’s lead author, described the new approach to human life as a sequence of events, similar to words forming sentences. He explained that they used transformer models, typically used in artificial intelligence, to analyze these “life sequences.”
The researchers used the model to estimate the likelihood of a person dying within the next four years.
The results matched current research, indicating that individuals in leadership positions or with higher incomes had better chances of survival. On the other hand, being male, having a skilled occupation, or being diagnosed with a mental health condition increased the risk of death.
Dr. Liemann stated, “Our primary focus is not the prediction itself but understanding the data elements that enable accurate forecasts like these.”
The model also demonstrated better accuracy in personality test results in a larger population segment compared to current artificial intelligence systems.
The researchers believe their framework can help uncover new mechanisms that affect life outcomes and design targeted interventions.
However, they caution against using this model for life insurance purposes due to ethical implications.
Dr. Liemann emphasized that using the model for insurance purposes would undermine the principle of mutual risk, which is essential in the insurance industry.
The team also highlighted ethical concerns regarding the use of Life2Vec, such as data protection, privacy, and bias.
They concluded that their research demonstrates the vast potential of these models, but their real-world application must be regulated to protect individuals’ rights.
