Bistro: A general purpose oracle for macroeconomic time series (www.bis.org)

Researchers have introduced the BIS Time-series Regression Oracle (BISTRO), a groundbreaking general-purpose model designed for macroeconomic forecasting. Unlike traditional econometric models that require specific tailoring for each task, BISTRO employs a transformer architecture similar to that of large language models (LLMs). It is fine-tuned on a vast repository of macroeconomic data from the Bank for International Settlements (BIS), enabling it to provide accurate forecasts across a variety of macroeconomic aggregates. Its ability to adapt to unfamiliar data patterns was demonstrated through successful predictions during the 2021 inflation surge, showcasing its impressive flexibility and zero-shot learning capabilities. BISTRO's significance lies in its potential to revolutionize economic forecasting and policy analysis. By eliminating the need for bespoke models for different forecasting tasks, it offers a more efficient and adaptable solution for central bankers and economists. Users can now explore scenario analyses by simply adjusting explanatory variables without the need to rebuild models. Accessible via Google Colab, BISTRO not only simplifies the forecasting process but also ensures users can take advantage of its advanced capabilities. This innovative approach marks a substantial shift in leveraging AI to address complex economic questions and facilitates a more effective framework for understanding evolving macroeconomic dynamics.