AI can predict life events with accuracy, scientists say

Artificial intelligence (AI) has taken a significant leap forward in predicting events in people’s lives, according to a recent scientific study. Scientists from renowned universities have developed a groundbreaking model that can accurately anticipate life events by analyzing vast amounts of data and utilizing language-processing AI algorithms.

The research project, conducted by a collaboration between DTU, University of Copenhagen, ITU, and Northeastern University, has shed light on the potential of AI to forecast future outcomes. By training sophisticated “transformer models,” similar to the ones used by ChatGPT, the researchers have successfully organized and analyzed data on residence, education, income, health, and working conditions.

The resulting model, called life2vec, outperformed other advanced neural networks in predicting crucial life events such as personality traits and even estimating the time of death. By encoding data into a mathematical structure known as vectors, life2vec efficiently organizes information related to birth, education, salary, housing, and health.

“We are fascinated by considering human life as a series of events, much like a sentence in a language consists of words. Our experiments demonstrate that transformer models in AI can be used to analyze life sequences, or the various events that occur throughout a person’s life,” explains Sune Lehmann, a professor at DTU and the lead author of the study.

However, the researchers also emphasize the ethical concerns that arise with such predictive models. Safeguarding sensitive data, preserving privacy, and addressing biases within the data are essential challenges that must be addressed before these models can be utilized in practical applications, such as assessing disease risks or predicting preventable life events.

Lehmann emphasizes the importance of engaging in a democratic conversation to discuss the impact of these technological advancements. “Similar technologies are already being used by tech companies to track our behavior on social networks and accurately profile us. It is crucial to discuss and consider the direction technology is taking us and whether this development aligns with our values,” he adds.

Moving forward, the researchers aim to incorporate additional types of data, such as text, images, and social connections, to enhance the accuracy of their predictions. This integration of data opens up new possibilities for collaboration between social and health sciences, paving the way for a deeper understanding of human behavior and its impact on future outcomes.

The study, titled “Using Sequences of Life-events to Predict Human Lives,” relied on comprehensive labor market data and information from the National Patient Registry and Statistics Denmark. These datasets encompassed the records of over 6 million Danes and included details on income, job type, industry, healthcare visits, diagnoses, and urgency levels.

Transformer models, which are advanced AI architectures, were employed in the study to process and understand language. These models have proven to be highly efficient, enabling the training of large language models on extensive datasets.