Would you put your trust in AI to make life predictions?

By analyzing health data, the artificial intelligence (AI) model life2vec can forecast life events, including the approximate time of death.

Researchers at DTU, the University of Copenhagen, ITU, and Northeastern University in the US have collaborated to develop a ground-breaking artificial intelligence (AI) model that leverages vast amounts of personal data to forecast life events, including the approximate time of death.

The project presents a state-of-the-art model called life2vec and is described in the recent Nature Computational Science paper titled “Using Sequences of Life-events to Predict Human Lives.”

AI forecast: What’s up with Life2vec?

the life2vec model was trained using a vast amount of health and labor market data from 6 million individuals. It is built on a transformer architecture, akin to that of OpenAI’s ChatGPT.

Following an initial learning phase, the model outperformed other state-of-the-art neural networks in terms of its prediction skills. It was able to predict personality traits and even the exact time of death with accuracy.

“We used the model to address the fundamental question: to what extent can we predict events in your future based on conditions and events in your past?” said Professor Sune Lehmann, the article’s primary author and a researcher at DTU. In terms of science, we are more interested in the features of the data that allow the model to produce such accurate results than we are in the forecast itself.

Forecasting events in life.

Predictions produced by life2vec provide interesting information like the probability of passing away within a given period of time. The model’s responses were analyzed, and the results were congruent with social science research already in existence. Notable correlations between higher wealth and survival chances were found, as well as leadership responsibilities.

It was discovered that there was a correlation between a higher risk of death and being male, skillful, or diagnosed with mental illness.

erratic happenings in life.

The researchers do, however, note that there are ethical questions with the life2vec model, such as privacy problems, safeguarding sensitive data, and possible biases in the data.

Before the model can be used, for example, to determine a person’s risk of getting an illness or going through avoidable life events, these issues must be fully resolved.

“Similar technologies for predicting life events and human behavior are already used inside tech companies that, for example, track our behavior on social networks, profile us extremely accurately, and use these profiles to predict our behavior and influence us,” Professor Lehmann said, emphasizing the significance of incorporating these ethical considerations into public discourse.

In order to improve the model’s prediction power, researchers intend to incorporate more data types in the future, including text, photos, and information about social connections.

Our comprehension of human life events could be completely transformed by this innovative use of data, which creates new opportunities for collaboration between the social and health sciences.
Drawing from health records and labor market data spanning from 2008 to 2020, the research offers an outlook for the future.

Leave a Comment