Artificial intelligence already can predict the weather or even the evolution of stock markets. From now on, it could also anticipate major events in our personal lives. This is the feat achieved by scientists through the development of an innovative AI system called Life2vec.
An AI system that uses data to make predictions
Generative AI systems such as Google’s Bard or OpenAI’s ChatGPT work on large predictive language models. These algorithms anticipate the most likely next word in a sentence, relying on the numerous elements of context assimilated during their training.
Intrigued by these predictive capabilities, Danish researchers explored the possibility of going further by using these models to predict major life events. To do this, they created their own generative AI, Life2vec, taking inspiration from how LLMs work. According to Sune Lehmann, lead author of the study and professor at the Technical University of Denmark: “ We can think of human life as a long sequence of events, in the same way that a sentence is a sequence of words .”
To develop their model, the researchers used a large amount of data. They collected the personal information of 6 million Danes over the period 2008-2020: income, health, age, profession, working hours, address, medical history, etc. It is therefore a real concentration of big data which will subsequently be transformed into an algorithmic crystal ball.
The next phase was to format this data in a way that made it usable by AI. The findings of this study were made public in the journal Nature Computational Science on December 19.
Highly precise results
After training on this data, the Life2vec model was subjected to prediction tests. The researchers selected a sample of people aged 35 to 65 in 2008, some of whom had died between 2016 and 2020. Then, they queried the AI system for information on which individuals were still alive as well as those who died during the period from 2016 to 2020.
The results of Life2vec are impressive. AI has proven capable of predicting future events in individuals’ lives with a high level of accuracy, far beyond the performance of other existing advanced algorithms. The system was able to predict things as complex as a person’s personality and the date of their death.
The researchers carefully examined the predictions generated by the Life2vec model. They found that its results reflected the conclusions of sociological analyses carried out in the past. For example, AI determined that higher socioeconomic status in 2008 was associated with greater longevity through 2020. Conversely, it identified precariousness and psychological disorders as factors of increased mortality over the period.
A promising tool, but challenges to overcome
According to the researchers, their study was not intended to predict mortality. Their objective is rather to explore a new methodology for analyzing the life trajectories of individuals. This technological advance opens the way to varied potential applications. For example, it could help in the development of public policies. Insurance companies would also find a risk assessment tool there .
However, researchers warn of crucial ethical considerations , including the protection of personal data used and the need to address algorithmic bias. According to them, this is very important before considering the use of the Life2vec model to predict aspects as sensitive as a person’s health or death.
In Conclusion
Life2vec has the potential to be a powerful tool for understanding ourselves, our health, and our place in the world. While its prediction of death might be the most sensationalized aspect, its capabilities extend far beyond, offering the potential to improve our lives and build a better future.
It’s important to remember that Life2vec’s predictions are not deterministic nor should they be considered absolute truths. They are simply tools to gain insights and make informed decisions. It’s always crucial to use critical thinking, exercise personal judgment, and consult with experts when necessary.
