Cutting-edge artificial intelligence designed to model written language has demonstrated an unprecedented capability: predicting significant life events and even estimating the time of death.
A collaborative research effort from DTU, University of Copenhagen, ITU, and Northeastern University in the US has harnessed the power of large-scale data to train transformer models, similar to ChatGPT, revealing their potential to systematically organize data and forecast pivotal moments in an individual’s life.
The study, titled “Using Sequences of Life-events to Predict Human Lives,” published in Nature Computational Science, delves into the analysis of health data and labor market attachment for a staggering 6 million Danes. The model, aptly named life2vec, surpassed other advanced neural networks in predicting outcomes, such as personality traits and the time of death, with remarkable accuracy after undergoing initial training to recognize patterns in the data.
Professor Sune Lehmann from DTU, the lead author of the article, emphasizes the broader implications of their research: “Scientifically, what is exciting for us is not so much the prediction itself, but the aspects of data that enable the model to provide such precise answers.”
Life2vec’s predictions extend to crucial questions like the likelihood of death within a specified timeframe. Intriguingly, the model’s responses align with existing social science findings, reinforcing associations between factors like leadership positions, higher income, and increased survival chances.
Lehmann elucidates, “What’s exciting is to consider human life as a long sequence of events, similar to how a sentence in a language consists of a series of words. This is usually the type of task for which transformer models in AI are used, but in our experiments, we use them to analyze what we call life sequences, i.e., events that have happened in human life.”
Data protection concerns
Despite the groundbreaking potential, ethical considerations loom large. Protecting sensitive data, preserving privacy, and addressing biases in the data are challenges that must be comprehensively tackled before models like life2vec can be deployed to assess risks or predict preventable life events.
Lehmann underscores the need for a democratic discourse around these technologies, stating, “This discussion needs to be part of the democratic conversation so that we consider where technology is taking us and whether this is a development we want.”
Looking ahead, the researchers envision incorporating additional information, such as text, images, or social connections, to further enhance the capabilities of data-driven models. The intersection of social and health sciences, driven by advanced data analysis, promises a new frontier in understanding and predicting human experiences.
