This study analyzes whether OpenAI's ChatGPT-3.5 and ChatGPT-4 can accurately predict future events using two response generation strategies: direct prediction and future narratives. To evaluate their accuracy, the limits of their training, which concluded in September 2021, were leveraged, and they were asked to predict events from 2022. It was found that the future narratives strategy significantly improved ChatGPT-4's accuracy, especially in predicting Oscar winners and economic trends, such as inflation and unemployment rates.
The study found that ChatGPT-4's ability to generate stories contributes to better data synthesis and extrapolation compared to direct prediction. Experiments were conducted with different approaches to examine whether applying narratives, such as fictional speeches by Federal Reserve Chairman Jerome Powell, could improve accuracy in economic prediction. Surprisingly, this method was observed to offer predictions close to actual consumer expectation surveys, highlighting an opportunity to leverage these models in forward-looking analyses.
These findings reinforce the potential of language models to generate valuable insights in areas such as economic planning and trend forecasting. In this context, Q2BSTUDIO, a leading company in development and technology services, explores and applies innovative solutions based on artificial intelligence and data analysis. Our expertise in software design and development allows us to integrate advanced predictive modeling technologies to improve business decision-making and optimize processes across various industries. With a vision focused on digital transformation, Q2BSTUDIO works on implementing technological strategies that maximize the potential of tools like ChatGPT for different market sectors.
