Researchers have developed a new artificial intelligence (AI) tool that leverages photographs to predict agricultural harvests. This innovative tool, named life2vec, uses sequences of life events—akin to the technology behind large language models like ChatGPT—to make its predictions.
The model, trained on a vast dataset from the entire population of Denmark, aims to create long patterns of recurring life events. This approach allows the AI to predict various outcomes, such as an individual’s lifespan, with higher accuracy than previous models (Northeastern Global News) (HT Tech).
One of the key aspects of life2vec is its ability to use data-driven insights to build vector representations in embedding spaces, which help categorise and connect life events like income, education, and health factors. These representations form the basis of the model’s predictive capabilities. For example, the tool can predict a person’s mortality probability by visualising data in a way that categorises individuals based on their likelihood of death (Gadgets 360).
Despite its potential, the researchers emphasise that this tool should not be used for real-life predictions. Instead, it should serve as a foundation for further research and development. “Even though we’re using prediction to evaluate how good these models are, the tool shouldn’t be used for prediction on real people,” said Professor Tina Eliassi-Rad from Northeastern University. This sentiment was echoed by Sune Lehmann, a co-author of the study and professor at the Technical University of Denmark, who highlighted the importance of considering the tool’s ethical implications and limitations (HT Tech) (MIT News).
Overall, life2vec represents a significant advancement in predictive modelling, offering a comprehensive reflection of human life events and their potential outcomes, paving the way for future developments in AI applications.
