Researchers from DTU, University of Copenhagen, ITU, and Northeastern University have developed artificial intelligence that can predict events in people’s lives. This AI uses large amounts of data to train transformer models, which can organize data and predict outcomes, including the time of death.
In their article, ‘Using Sequences of Life-events to Predict Human Lives’, published in Nature Computational Science, the researchers analyzed health and labor market data for 6 million Danes. The model, life2vec, outperformed other neural networks and accurately predicted outcomes such as personality and time of death.
The predictions from Life2vec are answers to general questions, such as predicting death within four years. The model encodes data in a system of vectors, organizing different data such as time of birth, schooling, income, and health. Ethical questions surrounding privacy, bias, and sensitive data have been raised by the researchers.
The researchers hope to incorporate text and image data and social connections. The use of data opens up new interactions between social and health sciences.
The dataset includes information on income, salary, job type, industry, and health records. Transformer models are used to train large language models on large datasets. Neural networks, including transformer models, rely on training data to improve their accuracy over time.