Artificial Intelligence (AI) has reached new heights, with researchers from DTU, University of Copenhagen, ITU, and Northeastern University unveiling a groundbreaking project that leverages AI to predict events in people’s lives. By harnessing the power of transformer models, similar to those used in ChatGPT, the team demonstrated the ability to systematically organize vast amounts of data about individuals, ultimately predicting outcomes ranging from personality traits to the timing of death.
Life2vec Model:
In their recent scientific article, titled “Using Sequences of Life-Events to Predict Human Lives,” published in Nature Computational Science, the researchers delve into their model, life2vec. This innovative approach involves analyzing health data and labor market engagement for a staggering 6 million Danes. The life2vec model, once trained, surpasses other advanced neural networks in accurately predicting life outcomes.
Predicting Personality and Time of Death:
After an initial training phase where the model learns intricate patterns within the data, life2vec stands out by predicting outcomes such as personality traits and even estimating the time of death with remarkable accuracy. The researchers emphasize that the model’s capability to make precise predictions raises questions not just about the predictions themselves, but also about the underlying data that enables such accuracy.
Life Sequences as Language:
Sune Lehmann, professor at DTU and the article’s first author, highlights the innovative aspect of considering human life as a sequence of events, akin to constructing a sentence in language. While transformer models are conventionally used for linguistic tasks, the researchers successfully employed them to analyze what they term “life sequences” — events that shape human lives.
Ethical Considerations:
Despite the promising results, the researchers acknowledge the ethical dilemmas associated with the life2vec model. Concerns about safeguarding sensitive data, ensuring privacy, and addressing biases in data must be comprehensively understood before considering the model for applications like assessing an individual’s risk of disease or preventable life events.
Sune Lehmann stresses the importance of a democratic conversation surrounding the implications of such technology. He notes that similar predictive technologies are already in use by tech companies, prompting a vital discussion about the trajectory of technology and its impact on society.
Future Directions:
Looking ahead, the researchers propose incorporating additional types of information, such as text, images, or social connections, to enhance the life2vec model. This multidimensional approach opens up new possibilities for interaction between social and health sciences, paving the way for a more comprehensive understanding of human life.
Conclusion:
As AI continues to advance, the life2vec model represents a significant stride in predicting human lives based on historical data. The ethical considerations surrounding its use underscore the need for a thoughtful and inclusive discourse about the trajectory of technology and its implications for society.
